text
stringlengths
70
7.94k
__index_level_0__
int64
105
711k
Title: Fusing Geometric and Scene Information for Cross-View Geo-Localization Abstract: ABSTRACTCross-view geo-localization is to match scene images (e.g. ground-view images) with geo-tagged aerial images, which is crucial to a wide range of applications such as autonomous driving and street view navigation. Existing methods can neither address the perspective difference well nor effectively capture the scene information. In this work, we propose a Geometric and Scene Information Fusion (GSIF) model for more accurate cross-view geo-localization. GSIF first learns the geometric information of scene images and aerial images via log-polar transformation and spatial-attention aggregation to alleviate the perspective difference. Then, it mines the scene information of scene images via Sky View Factor (SVF) extraction. Finally, both geometric information and scene information are fused for image matching, and a balanced loss function is introduced to boost the matching accuracy. Experimental results on two real datasets show that our model can significantly outperforms the existing methods.
710,623
Title: SpCQL: A Semantic Parsing Dataset for Converting Natural Language into Cypher Abstract: ABSTRACTThe Neo4j query language Cypher enables efficient querying for graphs and has become the most popular graph database language. Due to its complexities, semantic parsing (similar to Text-to-SQL) that translates natural language queries to Cypher becomes highly desirable. We propose the first Text-to-CQL dataset, SpCQL, which contains one Neo4j graph database, 10,000 manually annotated natural language queries and the matching Cypher queries (CQL). Correspondingly, based on this dataset, we define a new semantic parsing task Text-to-CQL. The Text-to-CQL task differs from the traditional Text-to-SQL task due to CQL being more flexible and versatile, especially for schema queries, which brings precedented challenges for the translation process. Although current SOTA Text-to-SQL models utilize SQL schema and contents, they do not scale up to large-scale graph databases. Besides, due to the absence of the primary and foreign keys in Cypher, which are essential for the multi-table Text-to-SQL task, existing Text-to-SQL models are rendered ineffective in this new task and have to be adapted to work. We propose three baselines based on the Seq2Seq framework and conduct experiments on the SpCQL dataset. The experiments yield undesirable results for existing models, hence pressing for subsequent research that considers the characteristics of SQL. The dataset is available at https://github.com/Guoaibo/Text-to-CQL.
710,624
Title: Binary Transformation Method for Multi-Label Stream Classification Abstract: ABSTRACTData streams produce extensive data with high throughput from various domains and require copious amounts of computational resources and energy. Many data streams are generated as multi-labeled and classifying this data is computationally demanding. Some of the most well-known methods for Multi-Label Stream Classification are Problem Transformation schemes; however, previous work on this area does not satisfy the efficiency demands of multi-label data streams. In this study, we propose a novel Problem Transformation method for Multi-Label Stream Classification called Binary Transformation, which utilizes regression algorithms by transforming the labels into a continuous value. We compare our method against three of the leading problem transformation methods using eight datasets. Our results show that Binary Transformation achieves statistically similar effectiveness and provides a much higher level of efficiency.
710,625
Title: End-to-end Multi-task Learning Framework for Spatio-Temporal Grounding in Video Corpus Abstract: ABSTRACTIn this paper, we consider a novel task, Video Corpus Spatio-Temporal Grounding (VCSTG) for material selection and spatio-temporal adaption in intelligent video editing. Given a text query depicting an object and a corpus of untrimmed and unsegmented videos, VCSTG aims to localize a sequence of spatio-temporal object tubes from the video corpus. Existing methods tackle the VCSTG task in a multi-stage approach, which encodes the query and video representation independently for each task, leading to local optimum. In this paper, we propose a novel one-stage multi-task learning based framework named MTSTG for the VCSTG task. MTSTG learns unified query and video representation for video retrieval, temporal grounding and spatial grounding tasks. Video-level, frame-level and object-level contrastive learning are introduced to measure the mutual information between query and video at different granularity. Comprehensive experiments demonstrate our newly proposed framework outperforms the state-of-the-art multi-stage methods on VidSTG dataset.
710,626
Title: MalNet: A Large-Scale Image Database of Malicious Software Abstract: ABSTRACTComputer vision is playing an increasingly important role in automated malware detection with the rise of the image-based binary representation. These binary images are fast to generate, require no feature engineering, and are resilient to popular obfuscation methods. Significant research has been conducted in this area, however, it has been restricted to small-scale or private datasets that only a few industry labs and research teams have access to. This lack of availability hinders examination of existing work, development of new research, and dissemination of ideas. We release MalNet-Image, the largest public cybersecurity image database, offering 24x more images and 70x more classes than existing databases (available at https://mal-net.org). MalNet-Image contains over 1.2 million malware images-across 47 types and 696 families---democratizing image-based malware capabilities by enabling researchers and practitioners to evaluate techniques that were previously reported in propriety settings. We report the first million-scale malware detection results on binary images. MalNet-Image unlocks new and unique opportunities to advance the frontiers of machine learning, enabling new research directions into vision-based cyber defenses, multi-class imbalanced classification, and interpretable security.
710,627
Title: On the Mining of Time Series Data Counterfactual Explanations using Barycenters Abstract: ABSTRACTEXplainable Artificial Intelligence (XAI) methods are increasingly accepted as effective tools to trace complex machine learning models' decision-making processes. There are two underlying XAI paradigms: (1) traditional factual methods and (2) emerging counterfactual models. The first family of methods uses feature attribution techniques that alter the feature space and observe the impact on the decision function. Counterfactual models aim at providing the smallest possible change to the feature vector that can change the prediction outcome. In this paper, we propose TimeX, a new model-agnostic time series counterfactual explanation algorithm that provides sparse, interpretable, and contiguous explanations. We validate our model using real-world time series datasets and show that our approach can generate explanations with up to 20% fewer outliers in comparison with other state-of-the-art competing baselines.
710,628
Title: Subspace Co-clustering with Two-Way Graph Convolution Abstract: ABSTRACTSubspace clustering aims to cluster high dimensional data lying in a union of low-dimensional subspaces. It has shown good results on the task of image clustering but text clustering, using document-term matrices, proved more impervious to advances based on this approach. We hypothesize that this is because, compared to image data, text data is generally higher dimensional and sparser. This renders subspace clustering impractical in such a context. Here, we leverage subspace clustering for text by addressing these issues. We first extend the concept of subspace clustering to co-clustering, which has been extensively used on document-term matrices due to the resulting interplay between the document and term representations. We then address the sparsity problem through a two-way graph convolution, which promotes the grouping effect that has been credited for the effectiveness of some subspace clustering models. The proposed formulation results in an algorithm that is efficient both in terms of computational and spatial complexity. We show the competitiveness of our model w.r.t the state-of-the-art on document-term attributed graph datasets in terms of performance and efficiency.
710,629
Title: Adaptive Graph Spatial-Temporal Transformer Network for Traffic Forecasting Abstract: ABSTRACTTraffic forecasting can be highly challenging due to complex spatial-temporal correlations and non-linear traffic patterns. Existing works mostly model such spatial-temporal dependencies by considering spatial correlations and temporal correlations separately, or within a sliding temporal window, and fail to model the direct spatial-temporal correlations. Inspired by the recent success of transformers in the graph domain, in this paper, we propose to directly model the cross-spatial-temporal correlations on the adaptive spatial-temporal graph using local multi-head self-attentions. We then propose a novel Adaptive Graph Spatial-Temporal Transformer Network (ASTTN), which stacks multiple spatial-temporal attention layers to apply self-attention on the input graph, followed by linear layers for predictions. Experimental results on public traffic network datasets, METR-LA PEMS-BAY, PeMSD4, and PeMSD7, demonstrate the superior performance of our model.
710,630
Title: Semi-Supervised Learning with Data Augmentation for Tabular Data Abstract: ABSTRACTData augmentation-based semi-supervised learning (SSL) methods have made great progress in computer vision and natural language processing areas. One of the most important factors is that the semantic structure invariance of these data allows the augmentation procedure (e.g., rotating images or masking words) to thoroughly utilize the enormous amount of unlabeled data. However, the tabular data does not possess an obvious invariant structure, and therefore similar data augmentation methods do not apply to it. To fill this gap, we present a simple yet efficient data augmentation method particular designed for tabular data and apply it to the SSL algorithm: SDAT (Semi-supervised learning with Data Augmentation for Tabular data). We adopt a multi-task learning framework that consists of two components: the data augmentation procedure and the consistency training procedure. The data augmentation procedure which perturbs in latent space employs a variational auto-encoder (VAE) to generate the reconstructed samples as augmented samples. The consistency training procedure constrains the predictions to be invariant between the augmented samples and the corresponding original samples. By sharing a representation network (encoder), we jointly train the two components to improve effectiveness and efficiency. Extensive experimental studies validate the effectiveness of the proposed method on the tabular datasets.
710,631
Title: MASR: A Model-Agnostic Sparse Routing Architecture for Arbitrary Order Feature Sharing in Multi-Task Learning Abstract: ABSTRACTMulti-task learning (MTL) has experienced rapid growth in recent years. A typical way of conducting MTL with deep neural networks (DNNs) is either establishing a sort of global feature sharing mechanism across all tasks or assigning each task an individual set of parameters with cross-connections. However, these existing approaches leverage DNNs only to share features of a certain order. Several modelsdemonstrated that explicitly modeling feature sharing with both low-order and high-order features can boost performance. To this end, we propose a model-agnostic sparse routing architecture called MASR, which emphasizes arbitrary order feature sharing for multi-task learning. It is able to choose specific orders of features to route for a given task through learnable latent variables. Moreover, MASR is model-agnostic and can be combined with existing MTL models to share features of both low-order and high-order. Extensive experimental results on several real-world datasets not only confirm the significant improvement of MASR performed to existing MTL models but also outperform existing hybrid architectures in terms of AUC metric.
710,632
Title: GFlow-FT: Pick a Child Network via Gradient Flow for Efficient Fine-Tuning in Recommendation Systems Abstract: ABSTRACTConversion Rate (CVR) prediction is a crucial task in online advertising systems. Existing single-domain CVR prediction models suffer from the data sparsity problem since few users purchase items after clicking. In recent years, a robust and effective technique called fine-tuning can transfer knowledge from a data-rich source domain to enhance the CVR prediction performance in a data-sparse target domain. However, since most CVR prediction models have a large number of parameters, fine-tuning all the parameters on a data-sparse domain may lead to over-fitting. In this paper, we propose a general and efficient transfer learning method called Gradient-Flow based Fine-Tuning (GFlow-FT), which only needs to update a subset of parameters (called child network) via pruning the gradients to restrain gradient norm against over-fitting. In addition, our method employs the gradient-flow based measure via calculating the Hessian-gradient product as the criteria for picking the child network, which is superior to the magnitude-based and loss-based measure from empirical results. Extensive experimental results on three real-world datasets from recommendation systems show that GFlow-FT can significantly improve the performance of CVR prediction compared with state-of-the-art fine-tuning approaches.
710,633
Title: OpeNTF: A Benchmark Library for Neural Team Formation Abstract: ABSTRACTWe contribute OpeNTF, an open-source python-based benchmark library to support neural team formation research. Team formation falls under social information retrieval (Social IR), where the right group of experts should be retrieved to solve a task, which is intractable due to the vast pool of feasible candidates with diverse skills. Even though neural networks could successfully address efficiency while maintaining efficacy, they lack standard implementation and experimental details, which calls for excessive efforts in repeating or reproducing the results in new domains. OpeNTF provides a standard and reproducible platform for neural team formation. It incorporates a host of canonical neural models along with three large-scale training datasets from varying domains. Leveraging an object-oriented structure, OpeNTF readily accommodates the addition of new neural models and training datasets. The first of its kind in neural team formation, OpeNTF also offers negative sampling heuristics that can be seamlessly integrated during model training to boost efficiency and to improve the effectiveness of inference.
710,634
Title: Effective Neural Team Formation via Negative Samples Abstract: ABSTRACTForming teams of experts who collectively hold a set of required skills and can successfully cooperate is challenging due to the vast pool of feasible candidates with diverse backgrounds, skills, and personalities. Neural models have been proposed to address scalability while maintaining efficacy by learning the distributions of experts and skills from successful teams in the past in order to recommend future teams. However, such models are prone to overfitting when training data suffers from a long-tailed distribution, i.e., few experts have most of the successful collaborations, and the majority has participated sparingly. In this paper, we present an optimization objective that leverages both successful and virtually unsuccessful teams to overcome the long-tailed distribution problem. We propose three negative sampling heuristics that can be seamlessly employed during the training of neural models. We study the synergistic effects of negative samples on the performance of neural models compared to lack thereof on two large-scale benchmark datasets of computer science publications and movies, respectively. Our experiments show that neural models that take unsuccessful teams (negative samples) into account are more efficient and effective in training and inference, respectively.
710,635
Title: LCD: Adaptive Label Correction for Denoising Music Recommendation Abstract: ABSTRACTMusic recommendation is usually modeled as a Click-Through Rate (CTR) prediction problem, which estimates the probability of a user listening a recommended song. CTR prediction can be formulated as a binary classification problem where the played songs are labeled as positive samples and the skipped songs are labeled as negative samples. However, such naively defined labels are noisy and biased in practice, causing inaccurate model predictions. In this work, we first identify serious label noise issues in an industrial music App, and then propose an adaptive Label Correction method for Denoising (LCD) music recommendation by ensembling the noisy labels and the model outputs to encourage a consensus prediction. Extensive offline experiments are conducted to evaluate the effectiveness of LCD on both industrial and public datasets. Furthermore, in a one-week online AB test, LCD also significantly increases both the music play count and time per user by 1% to 5%.
710,636
Title: A Dataset for Burned Area Delineation and Severity Estimation from Satellite Imagery Abstract: ABSTRACTThe ability to correctly identify areas damaged by forest wildfires is essential to plan and monitor the restoration process and estimate the environmental damages after such catastrophic events. The wide availability of satellite data, combined with the recent development of machine learning and deep learning methodologies applied to the computer vision field, makes it extremely interesting to apply the aforementioned techniques to the field of automatic burned area detection. One of the main issues in such a context is the limited amount of labeled data, especially in the context of semantic segmentation. In this paper, we introduce a publicly available dataset for the burned area detection problem for semantic segmentation. The dataset contains 73 satellite images of different forests damaged by wildfires across Europe with a resolution of up to 10m per pixel. Data were collected from the Sentinel-2 L2A satellite mission and the target labels were generated from the Copernicus Emergency Management Service (EMS) annotations, with five different severity levels, ranging from undamaged to completely destroyed. Finally, we report the benchmark values obtained by applying a Convolutional Neural Network on the proposed dataset to address the burned area identification problem.
710,637
Title: Dynamic Explicit Embedding Representation for Numerical Features in Deep CTR Prediction Abstract: ABSTRACTClick-Through Rate (CTR) prediction is a key problem in web search, recommendation systems, and online advertising display. Deep CTR models have achieved good performance due to adoption of the feature embedding and interaction. However, most research has focused on learning better feature interactions, with little attention to embedding representation. In this work, we propose a Dynamic Explicit Embedding Representation (DEER) for numerical features in deep CTR prediction, which can provide explicit and dynamic embedding representation for numerical features. The DEER framework is able to discretize numerical features automatically and dynamically, which can overcome the discontinuity problem in the representation of numeric information. Our methods are tested on two public datasets, and the experimental results show DEER can be applied to various deep CTR models, which also improve the performance effectively.
710,638
Title: CFS-MTL: A Causal Feature Selection Mechanism for Multi-task Learning via Pseudo-intervention Abstract: ABSTRACTMulti-task learning (MTL) has been successfully applied to a wide range of real-world applications. However, MTL models often suffer from performance degradation with negative transfer due to sharing all features without distinguishing their helpfulness for all tasks. To this end, many works on feature selection for multi-task learning (FS-MTL) have been proposed to alleviate negative transfer between tasks by learning features selectively for each specific task. However, due to latent confounders between features and task targets, the correlations captured by the feature selection modules proposed in these works may fail to reflect the actual effect of the features on the targets. This paper explains negative transfer in FS-MTL from a causal perspective and presents a novel architecture called Causal Feature Selection for Multi-task Learning(CFS-MTL). This method incorporates the idea of causal inference into feature selection for multi-task learning via pseudo-intervention. It aims to select features with more stable causal effects rather than spurious correlations for each task by regularizing the distance between feature ITEs and feature importance. We conduct extensive experiments based on three real-world datasets to demonstrate that our proposed CFS-MTL outperforms state-of-the-art MTL models significantly in the AUC metric.
710,639
Title: An Empirical Cross Domain-Specific Entity Recognition with Domain Vector Abstract: ABSTRACTRecognizing terminology entities across domains from professional texts is an important but challenging task in NLP. Most existing methods focus on recognizing generic entities, but few methods are to recognize the domain-specific entities across domains due to the very large discrepancy of entity representations between the source and target domains. To address this issue, we introduce domain vectors and context vectors to represent domain-specific semantics of entities and domain-irrelevant semantics of the context words, respectively. Based on the two types of vectors, we present a simple yet effective novel cross-domain named entity recognition approach, which aligns entity distributions between domains and separates entity distributions from context distributions for easily identifying entities. Experimental results demonstrate that the proposed approach can obtain significant improvement compared to existing cross-domain NER methods.
710,640
Title: Discriminative Language Model via Self-Teaching for Dense Retrieval Abstract: ABSTRACTDense retrieval (DR) has shown promising results in many information retrieval (IR) related tasks, whose foundation is high-quality text representations for effective search. Taking the pre-trained language models (PLMs) as the text encoders has become a popular choice in DR. However, the learned representations based on these PLMs often lose the discriminative power, and thus hurt the recall performance, particularly as PLMs consider too much content of the input texts. Therefore, in this work, we propose to pre-train a discriminative language representation model, called DiscBERT, for DR. The key idea is that a good text representation should be able to automatically keep those discriminative features that could well distinguish different texts from each other in the semantic space. Specifically, inspired by knowledge distillation, we employ a simple yet effective training method, called self-teaching, to distill the model's knowledge constructed when training on the sampled representative tokens of a text sequence into the model's knowledge for the entire text sequence. By further fine-tuning on publicly available retrieval benchmark datasets, DiscBERT can outperform the state-of-the-art retrieval methods.
710,641
Title: Deep Ordinal Neural Network for Length of Stay Estimation in the Intensive Care Units Abstract: ABSTRACTLength of Stay (LoS) estimation is important for efficient healthcare resource management. Since the distribution of LoS is highly skewed, some previous works frame the LoS estimation as a multi-class classification problem by dividing the range of LoS into buckets. However, they ignore the ordinal relationship between labels. The distribution of bucketed LoS, with a heavy head and a heavy tail, is still imbalanced since the long tail is grouped into the last bucket. This paper proposes a Deep Ordinal neural network for Length of stay Estimation in the intensive care units (DOSE). DOSE can exploit the ordinal relationship and mitigate the skewness. The ordinal classification problem is decomposed into a series of binary classification sub-problems by using multiple binary classifiers. To maintain consistency among binary classifiers, the monotonicity constraint penalty is proposed. The number of samples whose labels are higher or lower than a given threshold is at the same level due to the heavy head and tail of the distribution. Therefore, the training data of each binary classifier are balanced. Experiments are conducted on the real-world healthcare dataset. DOSE outperforms all baseline methods in all metrics. The distribution of the prediction of DOSE is more aligned with the ground truth.
710,642
Title: Marine-tree: A Large-scale Marine Organisms Dataset for Hierarchical Image Classification Abstract: ABSTRACTThis paper presents Marine-tree, a large-scale hierarchical annotated dataset for marine organism classification. Marine-tree contains more than 160k annotated images divided into 60 classes organised in a hierarchy-tree structure using an adapted CATAMI (Collaborative and Automated Tools for the Analysis of Marine Imagery and video) classification scheme. Images were meticulously collected by scuba divers using the RLS (Reef Life Survey) methodology and later annotated by experts in the field. We also propose a hierarchical loss function that can be applied to any multi-level hierarchical classification model, which takes into account the parent-child relationship between predictions and uses it to penalize inconsistent predictions. Experimental results demonstrate thatMarine-tree and the proposed hierarchical loss function are a good contribution for both research in underwater imagery and hierarchical classification.
710,643
Title: A Multi-Domain Benchmark for Personalized Search Evaluation Abstract: ABSTRACTPersonalization in Information Retrieval has been a hot topic in both academia and industry for the past two decades. However, there is still a lack of high-quality standard benchmark datasets for conducting offline comparative evaluations in this context. To mitigate this problem, in the past few years, approaches to derive synthetic datasets suited for evaluating Personalized Search models have been proposed. In this paper, we put forward a novel evaluation benchmark for Personalized Search with more than 18 million documents and 1.9 million queries across four domains. We present a detailed description of the benchmark construction procedure, highlighting its characteristics and challenges. We provide baseline performance including pre-trained neural models, opening room for the evaluation of personalized approaches, as well as domain adaptation and transfer learning scenarios. We make both datasets and models available for future research.
710,644
Title: IEEE13-AdvAttack A Novel Dataset for Benchmarking the Power of Adversarial Attacks against Fault Prediction Systems in Smart Electrical Grid Abstract: ABSTRACTDue to their economic and significant importance, fault detection tasks in intelligent electrical grids are vital. Although numerous smart grid (SG) applications, such as fault detection and load forecasting, have adopted data-driven approaches, the robustness and security of these data-driven algorithms have not been widely examined. One of the greatest obstacles in the research of the security of smart grids is the lack of publicly accessible datasets that permit testing the system's resilience against various types of assault. In this paper, we present IEEE13-AdvAttack, a large-scaled simulated dataset based on the IEEE-13 test node feeder suitable for supervised tasks under SG. The dataset includes both conventional and renewable energy resources. We examine the robustness of fault type classification and fault zone classification systems to adversarial attacks. Through the release of datasets, benchmarking, and assessment of smart grid failure prediction systems against adversarial assaults, we seek to encourage the implementation of machine-learned security models in the context of smart grids. The benchmarking data and code for fault prediction are made publicly available on https://bit.ly/3NT5jxG.
710,645
Title: Interpretability of BERT Latent Space through Knowledge Graphs Abstract: ABSTRACTThe advent of pretrained language have renovated the ways of handling natural languages, improving the quality of systems that rely on them. BERT played a crucial role in revolutionizing the Natural Language Processing (NLP) area. However, the deep learning framework it implements lacks interpretability. Thus, recent research efforts aimed to explain what BERT learns from the text sources exploited to pre-train its linguistic model. In this paper, we analyze the latent vector space resulting from the BERT context-aware word embeddings. We focus on assessing whether regions of the BERT vector space hold an explicit meaning attributable to a Knowledge Graph (KG). First, we prove the existence of explicitly meaningful areas through the Link Prediction (LP) task. Then, we demonstrate these regions being linked to explicit ontology concepts of a KG by learning classification patterns. To the best of our knowledge, this is the first attempt at interpreting the BERT learned linguistic knowledge through a KG relying on its pretrained context-aware word embeddings.
710,646
Title: Improving Imitation Learning by Merging Experts Trajectories Abstract: ABSTRACTThis paper proposes an original approach based on expert trajectories combination and Deep Reinforcement Learning to provide a better MineCraft player. The combination is based on the idea that the problem is naturally decomposable and the search space presents large plateaus. We use two steps approach to build a better trajectory from all existed expert trajectories and consequently to extract an optimal policy. The first step uses Birch clustering approach and images cosine similarity to obtain compact representation and substantial state and action space reduction. To reduce the overall complexity, the image distances are computed in images latent space trained by an encoder-decoder model. In the second step, we first eliminate plateaus to keep only the nodes with non-zero rewards then we compare trajectories using the Bellman equation and an appropriate value function. By checking the incremental compatibility of the trajectory of compact representations, we build the solution combining the best compatible sub-trajectories of the experts. The experimental results on NeurIPS MineRL 2020 challenge show that training the actors model on the most rewarding extracted subset of trajectories leads to achieve state-of-the-art performances on the MineCraft environment. The paper's source code is available here: https://github.com/thomJeffDoe/CompareTrajectories.
710,647
Title: Probing the Robustness of Pre-trained Language Models for Entity Matching Abstract: ABSTRACTThe paradigm of fine-tuning Pre-trained Language Models (PLMs) has been successful in Entity Matching (EM). Despite their remarkable performance, PLMs exhibit tendency to learn spurious correlations from training data. In this work, we aim at investigating whether PLM-based entity matching models can be trusted in real-world applications where data distribution is different from that of training. To this end, we design an evaluation benchmark to assess the robustness of EM models to facilitate their deployment in the real-world settings. Our assessments reveal that data imbalance in the training data is a key problem for robustness. We also find that data augmentation alone is not sufficient to make a model robust. As a remedy, we prescribe simple modifications that can improve the robustness of PLM-based EM models. Our experiments show that while yielding superior results for in-domain generalization, our proposed model significantly improves the model robustness, compared to state-of-the-art EM models.
710,648
Title: Scaling Up Mass-Based Clustering Abstract: ABSTRACTThis paper addresses the problem of scaling up the mass-based clustering paradigm to handle large datasets. The existing algorithm MBScan computes and stores all pairwise distances, resulting in quadratic time and space complexity. However, we observe that mass-based clustering requires information about only a tiny fraction of all possible data point pairs. We propose three optimizations to MBScan for quickly finding such pairs and computing their distances. We empirically evaluate our work on ten real-world and synthetic datasets. Our experiments show that our approach results in fast and memory-efficient clustering with no loss in the quality of clusters.
710,649
Title: Cross-Domain Product Search with Knowledge Graph Abstract: ABSTRACTThe notion personalization lies on the core of a real-world product search system, whose aim is to understand the user's search intent in a fine-grained level. The existing solutions mainly achieve this purpose through a coarse-grained semantic matching in terms of the query and item's description or the collective click correlations. Besides the issued query, the historical search behaviors of a user would cover lots of her personalized interests, which is a promising avenue to alleviate the semantic gap between users, items and queries. However, as to a specific domain, a user's search behaviors are generally sparse or even unavailable (i.e., cold-start users). How to exploit the search behaviors from the other relevant domain and enable effective fine-grained intent understanding remains largely unexplored for product search. Moreover, the semantic gap could be further aggravated since the properties of an item could evolve over time (e.g., the price adjustment for a mobile phone or the business plan update for a financial item), which is also mainly overlooked by the existing solutions. To this end, we are interested in bridging the semantic gap via a marriage between cross-domain transfer learning and knowledge graph. Specifically, we propose a simple yet effective knowledge graph based information propagation framework for cross-domain product search (named KIPS). In KIPS, we firstly utilize a shared knowledge graph relevant to both source and target domains as a semantic backbone to facilitate the information propagation across domains. Then, we build individual collaborative knowledge graphs to model both long-term interests/characteristics and short-term interests/characteristics of a user/item respectively. In order to harness cross-domain interest correlations, two unsupervised strategies to guide the interest learning and alignment are introduced: maximum mean discrepancy (MMD) and kg-aware contrastive learning. In detail, the MMD is utilized to support a coarse-grained domain alignment over the user's long-term interests across two domains. Then, the kg-aware contrastive learning process conducts a fine-grained interest alignment based on the shared knowledge graph. Experiments over two real-world large-scale datasets demonstrate the effectiveness of KIPS over a series of strong baselines. Our online A/B test also shows substantial performance gain on multiple metrics. Currently, KIPS has been deployed in AliPay for financial product search. Both the code implementation and the two datasets used for evaluation will be released online publicly.
710,650
Title: Breast Cancer Early Detection with Time Series Classification Abstract: ABSTRACTBreast cancer has become the leading cause of women cancer death worldwide. Despite the consensus that breast cancer early detection can significantly reduce treatment difficulty and cancer mortality, people still are reluctant to go to hospital for regular checkups due to the high costs incurred. A timely, private, affordable, and effective household breast cancer early detection solution is badly needed. In this paper, we propose a household solution that utilizes pairs of sensors embedded in the bra to measure the thermal and moisture time series data (BTMTSD) of the breast surface and conduct time series classification (TSC) to diagnose breast cancer. Three main challenges are encountered when doing BTMTSD classification, (1) small supervised dataset, which is a common limitation of medical research, (2) noisy time series with unique noise patterns, and (3) complex interplay patterns across multiple time series dimensions. To mitigate these problems, we incorporate multiple data augmentation and transformation techniques with various deep learning TSC approaches and compare their performances for the BTMTSD classification task. Experimental results validate the effectiveness of our framework in providing reliable breast cancer early detection.
710,651
Title: SASNet: Stage-aware Sequential Matching for Online Travel Recommendation Abstract: ABSTRACTSequential matching, which aims to predict the item a user will next interact with in the sequential context of the user's historical behaviors, is widely adopted in recommender systems. Existing works mainly characterize the sequential context as the dependencies of user interactions, which is less effective for online travel recommendation where users' behaviors are highly correlated with theirstages in the travel life cycle. Specifically, users on an online travel platform (OTP) usually go through different stages (e.g., exploring a destination, planning an itinerary), and make several correlated interactions (e.g., booking a flight, reserving a hotel, renting a car) at each stage. In this paper, we propose to capture the deep sequential context by modeling the evolving of user stages, and develop a novel stage-aware deep sequential matching network (SASNet) that incorporates inter-stage and intra-stage dependencies over stage-augmented interaction sequence for more accurate and interpretable recommendation. Extensive experiments on real-world datasets validate the superiority of our model for both online travel recommendation and general next-item recommendation. Our model has been successfully deployed at Fliggy, one of the most popular OTPs in China, and shows good performance in serving online traffic.
710,652
Title: Towards Edge-Cloud Collaborative Machine Learning: A Quality-aware Task Partition Framework Abstract: ABSTRACTEdge-cloud collaborative tasks with real-world services emerge in recent years and attract worldwide attention. Unfortunately, state-of-the-art edge-cloud collaborative machine-learning services are still not that reliable due to the data heterogeneity on the edge, where we usually have access to a mixed-up training set, which is intrinsically collected from various distributions of underlying tasks. Finding such hidden tasks that need to be revealed from given datasets is called the Task Partition problem. Manual task partition is usually expensive, unscalable, and biased. Accordingly, we propose Quality-aware Task Partition (QTP) problem, in which final tasks are partitioned by the performance of task models. To the best of our knowledge, this work is the first one to study the QTP problem with an emphasis on task quality. We also implement a public service, HiLens on Huawei Cloud, to support the whole process. We develop a polynomial-time algorithm namely the Task-Forest algorithm (TForest). TForest shows its superiority based on a case study with 57 real-world cameras. Compared with STOA baselines, TForest has on average 9.2% higher F1-scores and requires 43.1% fewer samples when deploying new cameras. Partial code of the framework has been adopted and released to KubeEdge-Sedna.
710,653
Title: Graph-based Weakly Supervised Framework for Semantic Relevance Learning in E-commerce Abstract: ABSTRACTProduct searching is fundamental in online e-commerce systems, it needs to quickly and accurately find the products that users required. Relevance is essential for e-commerce search, which role is avoiding displaying products that do not match search intent and optimizing user experience. Measuring semantic relevance is necessary because distributional biases between search queries and product titles may lead to large lexical differences between relevant textual expressions. Several problems limit the performance of semantic relevance learning, including extremely long-tail product distribution and low-quality labeled data. Recent works attempt to conduct relevance learning through user behaviors. However, noisy user behavior can easily cause inadequately semantic modeling. Therefore, it is valuable but challenging to utilize user behavior in relevance learning. In this paper, we first propose a weakly supervised contrastive learning framework that focuses on how to provide effective semantic supervision and generate reasonable representation. We utilize topology structure information contained in a user behavior heterogeneous graph to design a semantically aware data construction strategy. Besides, we propose a contrastive learning framework suitable for e-commerce scenarios with targeted improvements in data augmentation and training objectives. For relevance calculation, we propose a novel hybrid method that combines fine-tuning and transfer learning. It eliminates the negative impacts caused by distributional bias and guarantees semantic matching capabilities. Extensive experiments and analyses show the promising performance of proposed methods in relevance learning.
710,654
Title: Hierarchical Reinforcement Learning using Gaussian Random Trajectory Generation in Autonomous Furniture Assembly Abstract: ABSTRACTIn this paper, we propose a Gaussian Random Trajectory guided Hierarchical Reinforcement Learning (GRT-HL) method for autonomous furniture assembly. The furniture assembly problem is formulated as a comprehensive human-like long-horizon manipulation task that requires a long-term planning and a sophisticated control. Our proposed model, GRT-HL, draws inspirations from the semi-supervised adversarial autoencoders, and learns latent representations of the position trajectories of the end-effector. The high-level policy generates an optimal trajectory for furniture assembly, considering the structural limitations of the robotic agents. Given the trajectory drawn from the high-level policy, the low-level policy makes a plan and controls the end-effector. We first evaluate the performance of GRT-HL compared to the state-of-the-art reinforcement learning methods in furniture assembly tasks. We demonstrate that GRT-HL successfully solves the long-horizon problem with extremely sparse rewards by generating the trajectory for planning.
710,655
Title: An Actor-critic Reinforcement Learning Model for Optimal Bidding in Online Display Advertising Abstract: ABSTRACTThe real-time bidding (RTB) paradigm allows the advertisers to submit a bid for each impression in online display advertising. A usual demand of the advertisers is to maximize the total value of winning impressions under constraints on some key performance indicators. Unfortunately, the existing RTB research in industrial applications can hardly achieve the optimum due to the stochastic decision scenarios and complex consumer behaviors. In this study, we address the application of RTB to mobile gaming where the in-app purchase action is of high uncertainty, making it challenging to evaluate individual impression opportunities. We first formulate the bidding process into a constrained optimization problem and then propose an actor-critic reinforcement learning (ACRL) model for obtaining the optimal policy under a dynamic decision environment. To avoid feeding too many samples with zero labels to the model, we provide a new way to quantify impression opportunities by integrating the in-app actions, such as conversion and purchase, and the characteristics of the candidate ad inventories. Moreover, the proposed ACRL learns a Gaussian distribution to simulate the audience's decision in a more real bidding scenario by taking additional contextual side information about both media and the audience. We also introduce how to deploy the learned model online to help adjust the final bid. At last, we conduct comprehensive offline experiments to demonstrate the effectiveness of ACRL and carefully set an online A/B testing experiment. The online experimental results verify the efficacy of the proposed ACRL in terms of multiple critical commercial indicators. ACRL has been deployed in the Tencent online display advertising platform and impacts billions of traffic every day. We believe proposed modifications for optimal bidding problems in RTB are practically innovative and can inspire the relative works in this field.
710,656
Title: Temporal and Heterogeneous Graph Neural Network for Financial Time Series Prediction Abstract: ABSTRACTThe price movement prediction of stock market has been a classical yet challenging problem, with the attention of both economists and computer scientists. In recent years, graph neural network has significantly improved the prediction performance by employing deep learning on company relations. However, existing relation graphs are usually constructed by handcraft human labeling or nature language processing, which are suffering from heavy resource requirement and low accuracy. Besides, they cannot effectively response to the dynamic changes in relation graphs. Therefore, in this paper, we propose a temporal and heterogeneous graph neural network-based (THGNN) approach to learn the dynamic relations among price movements in financial time series. In particular, we first generate the company relation graph for each trading day according to their historic price. Then we leverage a transformer encoder to encode the price movement information into temporal representations. Afterward, we propose a heterogeneous graph attention network to jointly optimize the embeddings of the financial time series data by transformer encoder and infer the probability of target movements. Finally, we conduct extensive experiments on the stock market in the United States and China. The results demonstrate the effectiveness and superior performance of our proposed methods compared with state-of-the-art baselines. Moreover, we also deploy the proposed THGNN in a real-world quantitative algorithm trading system, the accumulated portfolio return obtained by our method significantly outperforms other baselines.
710,657
Title: DuTraffic: Live Traffic Condition Prediction with Trajectory Data and Street Views at Baidu Maps Abstract: ABSTRACTThe task of live traffic condition prediction, which aims at predicting live traffic conditions (i.e., fast, slow, and congested) based on traffic information on roads, plays a vital role in intelligent transportation systems, such as navigation, route planning, and ride-hailing services. Existing solutions have adopted aggregated trajectory data to generate traffic estimates, which inevitably suffer from GPS drift caused by cluttered urban road scenarios. In addition, the trajectory information alone is insufficient to provide evidence for sudden traffic situations and perception of street-wise elements. To alleviate these problems, in this paper, we present DuTraffic, which is a robust and production-ready solution for live traffic condition prediction by taking both trajectory data and street views into account. Specifically, the vision-based detection and segmentation modules are developed to forecast traffic flow by using street views. Then, we propose a spatial-temporal-based module, TRST-Net, to learn the latent trajectory representation. Finally, a bilinear model is introduced to mix these two representations and then predicts live traffic conditions with trajectory data and street views in a mutually complementary manner. The task is recast as a multi-task learning problem, which could benefit from the strong representation of latent space manifold modeling. Extensive experiments conducted on large-scale, real-world datasets from Baidu Maps demonstrate the superiority and effectiveness of DuTraffic. In addition, DuTraffic has already been deployed in production at Baidu Maps since December 2020, handling tens of millions of requests every day. This demonstrates that DuTraffic is a practical and robust industrial solution for live traffic condition prediction.
710,658
Title: DuARUS: Automatic Geo-object Change Detection with Street-view Imagery for Updating Road Database at Baidu Maps Abstract: ABSTRACTAs the core foundation of web mapping, each geographic object (geo-object), such as a traffic sign, plays a vital role in navigation and intelligent driving. Determining how to obtain the latest high-precision geo-object information is a classic topic in updating road databases. Benefiting from the cost-effective attribute and availability of the positioning equipment and camera, the vision-based update pattern is becoming increasingly popular in the industry. Generally speaking, the road database update mainly includes three phases: geo-object recognition, localization, and change detection. Previous change detection strategies are mainly performed by comparing the historical road information (i.e., geo-object type and position) with the new geographic data of geo-objects collected from the street-view imagery. However, limited by the localization precision of the positioning equipment and the discriminative power of the vanilla differential-based method, the accuracy, recall, and efficiency of previous systems for geo-object change detection are greatly impaired. In addition, the artificially prescribed production standards make the geo-object position in the map data deviate from its position in the real world, as well as some geo-objects do not need to be updated (e.g., temporary speed limit), which further yields many false-positive detections and significantly increases the labor costs of existing systems. To address these challenges, we propose a novel framework called DuARUS for automatic geo-object change detection with street-view imagery. In this paper, we mainly focus on automatic geo-object localization and change detection. Specifically, for geo-object localization, we propose a two-stage, integrated localization algorithm based on image matching and monocular depth estimation. Furthermore, to achieve automatic change detection, vision-based representation learning and scene understanding strategies are introduced to build a large-scale geo-object semantic map, which can provide sufficient multimodal information support for change detection. Based on such artful modeling, we recast the complicated, labor-based change detection problem as a vanilla binary classification task, which is a robust and efficient strategy that contributes to resolving this problem. By combining these operations, we construct an industrial-grade, fully automatic production system for road database updates. Extensive experiments conducted on large-scale, real-world datasets from Baidu Maps demonstrate the superiority and effectiveness of the system. Moreover, this system has already been deployed in production at Baidu Maps since July 2020, handling 96% of automatic road database updates. DuARUS improves the annual update mileage from millions to tens of millions, and it achieves weekly updates.
710,659
Title: CTRL: Cooperative Traffic Tolling via Reinforcement Learning Abstract: ABSTRACTPeople have been working long to tackle the traffic congestion problem. Among the different measures, traffic tolling has been recognized as an effective way to mitigate citywide congestion. However, traditional tolling methods can not deal with the dynamic traffic flow in cities. Meanwhile, thanks to the development of traffic sensing technology, how to set appropriate dynamic tolling according to real time traffic observations has attracted research attention in recent years. In this paper, we put the dynamic tolling problem in a reinforcement learning setting and try to tackle the three key challenges of complex state representation, pricing action credit assignment, and route price relative competition. We propose a soft actor-critic method with (1) a route-level state attention, (2) an interpretable and provable reward design, and (3) a competition-aware Q attention. Extensive experiments on real datasets have shown the superior performance of our proposed method. In addition, interesting analysis on pricing actions and vehicle routes have demonstrated why the proposed method can outperform baselines.
710,660
Title: A Dual Channel Intent Evolution Network for Predicting Period-Aware Travel Intentions at Fliggy Abstract: ABSTRACTFliggy of Alibaba group is one of the largest online travel platform (OTPs) in China, which provides travel products and travel experiences for tens of millions of online users by the personalized recommendation system (RS). User's future travel intent prediction is one key problem in travel scenario, which decides where and what to recommend, e.g., traveling to a surrounding city or a distant city. Such travel intent prediction problem has a lot of important applications, e.g., to push a notification with surrounding scenic spots recommendation to a user with intent to travel around, or to enable personalized promotion strategies to users with different intents. Existing studies on user's intent are largely sub-optimal for users' travel intent prediction at OTPs, since they rarely pay attentions to the characteristics of the travel industry, namely, user behavior sparsity due to low frequency of travel, spatial-temporal periodicity patterns, and the correlations between user's online and offline behaviors. In this paper, to address these challenges, we propose a dual channel intent evolution network based online-offline periodicity-aware network, DCIEN, for user's future travel intent prediction. In particular, it consists of two basic components including 1) Spatial-temporal Intent Patterns Network(ST-IPN), which exploits users' periodic intent patterns from offline data based on convolutional neural networks; 2) Periodicity-aware Intent Evolution Network(PA-IEN), which captures user's instant intent from online behaviors data and the interactions between online and offline intents. Extensive offline and online experiments on a real-world OTP demonstrate the superior performance of DCIEN over state-of-the-art methods.
710,661
Title: WARNER: Weakly-Supervised Neural Network to Identify Eviction Filing Hotspots in the Absence of Court Records Abstract: ABSTRACTThe widespread eviction of tenants across the United States has metamorphosed into a challenging public-policy problem. In particular, eviction exacerbates several income-based, educational, and health inequities in society, e.g., eviction disproportionately affects low-income renting families, many of whom belong to underrepresented minority groups. Despite growing interest in understanding and mitigating the eviction crisis, there are several legal and infrastructural obstacles to data acquisition at scale that limit our understanding of the distribution of eviction across the United States. To circumvent existing challenges in data acquisition, we propose WARNER, a novel Machine Learning (ML) framework that predicts eviction filing hotspots in US counties from unlabeled satellite imagery dataset. We account for the lack of labeled training data in this domain by leveraging sociological insights to propose a novel approach to generate probabilistic labels for a subset of an unlabeled dataset of satellite imagery, which is then used to train a neural network model to identify eviction filing hotspots. Our experimental results show that WARNER acheives a higher predictive performance than several strong baselines. Further, the superiority of WARNER can be generalized to different counties across the United States. Our proposed framework has the potential to assist NGOs and policymakers in designing well-informed (data-driven) resource allocation plans to improve the nationwide housing stability. This work is conducted in collaboration with The Child Poverty Action Lab (a leading non-profit leveraging data-driven approaches to inform actions for relieving poverty and relevant problems in Dallas County, TX). The code can be accessed via https://github.com/maryam-tabar/WARNER.
710,662
Title: Sub-Task Imputation via Self-Labelling to Train Image Moderation Models on Sparse Noisy Data Abstract: ABSTRACTE-commerce marketplaces protect shopper experience and trust at scale by deploying deep learning models trained on human annotated moderation data, for the identification and removal of advert imagery that does not comply with moderation policies (a.k.a. defective images). However, human moderation labels can be hard to source for smaller advert programs that target specific device types with separate formats or for recently launched locales with unique moderation policies. Additionally, the sourced labels can be noisy due to annotator biases or policy rules clubbing multiple types of transgressions into a single category. Therefore, training advert image moderation models necessitates an approach that can effectively improve the sample efficiency of training, weed out noise and discover latent moderation sub-labels in one go. Our work demonstrates the merits of automated sub-label discovery using self-labelling. We show that self-labelling approaches can be used to decompose an image moderation task into its hidden sub-tasks (corresponding to intercepting a single sub-label) in an unsupervised manner, thus helping with cases where the granularity of labels is inadequate. This enables us to bootstrap useful representations quickly, via low-capacity but fast-learning teacher models that each specialize in a single distinct sub-task of the main classification task. These sub-task specialists then distil their logits to a high-capacity but slow-learning generalist student model, thus allowing it to perform well on complex moderation tasks with relatively fewer labels than vanilla supervised training. We conduct all our experiments on the moderation of sexually explicit advert images (though this method can be utilized for any defect type) and show a sizeable improvement in NPV (+30.2% absolute gain) viz-a-viz regular supervised baselines at a 1% FPR level. A long-term A/B test of our deployed model shows a significant relative reduction (-45.6%) in the prevalence of such advertisements compared to the previously deployed model.
710,663
Title: High Availability Framework and Query Fault Tolerance for Hybrid Distributed Database Systems Abstract: ABSTRACTModern commercial database systems are increasingly evolving into a hybrid distributed system model where a primary database host system enlists the services of a loosely coupled secondary system that acts as an accelerator. Often the secondary system is a distributed system that can perform specific tasks massively parallelized with results fed back to the host database. Similar models can also be seen in architectures that separate compute from storage. As the scale of the system grows, failures of nodes become common, and the architectural goal is to recover the system with minimal disruption to the workload as seen by the user. This paper introduces a new framework that allows a host database to efficiently manage the availability of a massive secondary distributed system and describes a mechanism to achieve query fault tolerance at the primary database by transparently re-executing query (sub)plans on the secondary distributed system. The focus is on improving two important aspects of disruption ? downtime and transparency to the user. The proposed mechanisms achieve quick recovery, reduced duration of downtime and isolation of errors during query execution, thus improving execution transparency for the users.
710,664
Title: Observability of SQL Hints in Oracle Abstract: ABSTRACTObservability is a critical requirement of increasingly complex and cloud-first data management systems. In most commercial databases, this relies on telemetry like logs, traces, and metrics, which helps to identify, mitigate, and resolve issues expeditiously. SQL monitoring tools, for example, can show how a query is performing. One area that has received comparatively less attention is the observability of the query optimizer whose inner workings are often shrouded in mystery. Optimizer traces can illuminate the plan selection process for a query, but they are comprehensible only to human experts and are not easily machine-parsable to remediate sub-optimal plans. Hints are directives that guide the optimizer toward specific directions. While hints can be used manually, they are often used by automatic SQL plan management tools that can quickly identify and resolve regressions by selecting alternate plans. It is important to know when input hints are inapplicable so that the tools can try other strategies. For example, a manual hint may have syntax errors, or an index in an automatic hint may have been accidentally dropped. In this paper, we describe the design and implementation of Oracle's hint observability framework which provides a comprehensive usage report of all hints, manual or otherwise, used to compile a query. The report, which is available directly in the execution plan in a human-understandable and machine-readable format, can be used to automate any necessary corrective actions. This feature is available in Oracle Autonomous Database 19c.
710,665
Title: Learning-to-Spell: Weak Supervision based Query Correction in E-Commerce Search with Small Strong Labels Abstract: ABSTRACTFor an E-commerce search engine, users finding the right product critically depend on spell correction. A misspelled query can fetch totally unrelated results which in turn leads to a bad customer experience. Around 32% of queries have spelling mistakes on our e-commerce search engine. The spell problem becomes more challenging when most spell errors arise from customers with little or no exposure to the English language besides the usual source of accidental mistyping on keyboard. These spell errors are heavily influenced by the colloquial and spoken accents of the customers. This limits the benefit from using generic spell correction systems which are learnt from cleaner English sources like Brown Corpus and Wikipedia with a very low focus on phonetic/vernacular spell errors. In this work, we present a novel approach towards spell correction that effectively solves a very diverse set of spell errors and outperforms several state-of-the-art systems in the domain of E-commerce search. Our strategy combines Learning-to-Rank on a small strongly labelled data with multiple learners trained with weakly labelled data. We report the effectiveness of our solution WellSpell (Weak and strong Labels for Learning to Spell) with both the offline evaluations and online A/B experiment.
710,666
Title: Multimodal Meta-Learning for Cold-Start Sequential Recommendation Abstract: ABSTRACTIn this paper, we study the task of cold-start sequential recommendation, where new users with very short interaction sequences come with time. We cast this problem as a few-shot learning problem and adopt a meta-learning approach to developing our solution. For our task, a major obstacle of effective knowledge transfer that is there exists significant characteristic divergence between old and new interaction sequences for meta-learning. To address the above issues, we purpose a Multimodal MetaLearning (denoted as MML) approach that incorporates multimodal side information of items (e.g., text and image) into the meta-learning process, to stabilize and improve the meta-learning process for cold-start sequential recommendation. In specific, we design a group of multimodal meta-learners corresponding to each kind of modality, where ID features are used to develop the main meta-learner and the rest text and image features are used to develop auxiliary meta-learners. Instead of simply combing the predictions from different meta-learners, we design an adaptive, learnable fusion layer to integrate the predictions based on different modalities. Meanwhile, we design a cold-start item embedding generator, which utilize multimodal side information to warm up the ID embeddings of new items. Extensive offline and online experiments demonstrate that MML can significantly improve the recommendation performance for cold-start users compared with baseline models. Our code is released at https://github.com/RUCAIBox/MML.
710,667
Title: Guided Text-based Item Exploration Abstract: ABSTRACTExploratory Data Analysis (EDA) provides guidance to users to help them refine their needs and find items of interest in large volumes of structured data. In this paper, we develop GUIDES, a framework for guided Text-based Item Exploration (TIE). TIE raises new challenges: (i) the need to abstract and query textual data and (ii) the need to combine queries on both structured and unstructured content. GUIDES represents text dimensions such as sentiment and topics, and introduces new text-based operators that are seamlessly integrated with traditional EDA operators. To train TIE policies, it relies on a multi-reward function that captures different textual dimensions, and extends the Deep Q-Networks (DQN) architecture with multi-objective optimization. Our experiments on Amazon and IMDb, two real-world datasets, demonstrate the necessity of capturing fine-grained text dimensions, the superiority of using both text-based and attribute-based operators over attribute-based operators only, and the need for multi-objective optimization.
710,668
Title: MIC: Model-agnostic Integrated Cross-channel Recommender Abstract: ABSTRACTSemantically connecting users and items is a fundamental problem for the matching stage of an industrial recommender system. Recent advances in this topic are based on multi-channel retrieval to efficiently measure users' interest on items from the massive candidate pool. However, existing studies are primarily built upon pre-defined retrieval channels, including User-CF (U2U), Item-CF (I2I), and Embedding-based Retrieval (U2I), thus access to the limited correlation between users and items which solely entail from partial information of latent interactions. In this paper, we propose a model-agnostic integrated cross-channel (MIC) approach for the large-scale recommendation, which maximally leverages the inherent multi-channel mutual information to enhance the matching performance. Specifically, MIC robustly models correlation within user-item, user-user, and item-item from latent interactions in a universal schema. For each channel, MIC naturally aligns pairs with semantic similarity and distinguishes them otherwise with more uniform anisotropic representation space. While state-of-the-art methods require specific architectural design, MIC intuitively considers them as a whole by enabling the complete information flow among users and items. Thus MIC can be easily plugged into other retrieval recommender systems. Extensive experiments show that our MIC helps several state-of-the-art models boost their performance on four real-world benchmarks. The satisfactory deployment of the proposed MIC on industrial online services empirically proves its scalability and flexibility.
710,669
Title: MEMENTO: Neural Model for Estimating Individual Treatment Effects for Multiple Treatments Abstract: ABSTRACTLearning individual level treatment effects from observational data is a problem of growing interest. For instance, inferring the effect of delivery promises on purchase of products on an e-commerce site or selecting the most effective treatment for a specific patient. Although the scenarios where we want to estimate the treatment effects in presence of multiple treatments is quite common in real life, most existing works related to individual treatment effect (ITE) are focused primarily on binary treatments and do not have a natural extension to the multi-treatment scenarios. In this paper we present MEMENTO ? a methodology and a framework to estimate individual treatment effect for multi-treatment scenarios, where the treatments are discrete and finite. Our approach is based on obtaining matching representations of the confounders for the various treatment types. This is achieved through minimization of an upper bound on the sum of factual and counterfactual losses. Experiments on real and semi-synthetic datasets show that MEMENTO is able to outperform known techniques for multi-treatment scenarios by close to 10% in certain use-cases. The proposed framework has been deployed for the problem of identifying minimum order quantity of a product in Amazon in an emerging marketplace and has re- sulted in a 4.7% reduction in shipping costs as proved from an A/B experiment.
710,670
Title: Efficient Compression Method for Roadside LiDAR Data Abstract: ABSTRACTRoadside LiDAR (Light Detection and Ranging) sensors are recently being explored for intelligent transportation systems aiming at safer and faster traffic management and vehicular operations. A key challenge in such systems is to efficiently transfer massive point-cloud data from the roadside LiDAR devices to the edge connected through a 5G network for real-time processing. In this paper, we consider the problem of compressing roadside (i.e. static) LiDAR data in real-time that provides a unique condition unexplored by current methods. Existing point-cloud compression methods assume moving LiDARs (that are mounted on vehicles) and do not exploit spatial consistency across frames over time. To this end, we develop a novel grouped wavelet technique for static roadside LiDAR data compression (i.e. SLiC). Our method compresses LiDAR data both spatially and temporally using a kd-tree data structure based on Haar wavelet coefficients. Experimental results show that SLiC can compress up to 1.9× more effectively than the state-of-the-art compression method can do. Moreover, SLiC is computationally more efficient to achieve 2× improvement in bandwidth usage over the best alternative. Even with this impressive gain in communication and storage efficiency, SLiC retains down-the-pipeline application's accuracy.
710,671
Title: Towards Fair Workload Assessment via Homogeneous Order Grouping in Last-mile Delivery Abstract: ABSTRACTThe popularity of e-commerce has promoted the rapid development of the logistics industry in recent years. As an important step in logistics, last-mile delivery from delivery stations to customers' addresses is now mainly finished by couriers, which requires accurate workload assessment based on actual efforts. However, the state-of-the-practice assessment methods neglect a vital factor that orders with the same customer's address (i.e., Homogeneous orders) can be delivered in a group (i.e., in a single trip) or separately (i.e., in multiple trips). It would cause unfair assessment among couriers if following the same rule. Thus, grouping homogeneous order accurately in the workload assessment is significant for achieving fair courier's workload assessment. To this end, we design, implement, and deploy a nationwide homogeneous order grouping system called FHOG for improving the accuracy of homogeneous order grouping in last-mile delivery for fair courier's workload assessment. FHOG utilizes the courier's reporting behavior for order inspection, collection, and delivery to identify homogeneous orders in the delivery station simultaneously for homogeneous order grouping. Compared with the state-of-the-practice method, our evaluation shows FHOG can effectively reduce order amounts with the higher and lower assessed courier's workload. We further deploy FHOG online in 8336 delivery stations to provide homogeneous order grouping service for more than 120 thousand couriers and 12 million daily orders. The results of the two surveys show that the couriers' acceptance rate is improved by 67% with FHOG after the promotion.
710,672
Title: STARDOM: Semantic Aware Deep Hierarchical Forecasting Model for Search Traffic Prediction Abstract: ABSTRACTWe study the search traffic forecasting problem for guaranteed search advertising (GSA) application in e-commerce platforms. The consumers express their purchase intents by posing queries to the e-commerce search engine. GSA is a type of guaranteed delivery (GD) advertising strategy, which forecasts the traffic of search queries, and charges the advertisers according to the predicted volumes of search queries the advertisers willing to buy. We employ the time series forecasting method to make the search traffic prediction. Different from existing time series prediction methods, search queries are semantically meaningful, with semantically similar queries possessing similar time series. And they can be grouped according to the brands or categories they belong to, exhibiting hierarchical structures. To fully take advantage of these characteristics, we design a SemanTic AwaRe Deep hierarchical fOrecasting Model (STARDOM for short) which explores the queries' semantic information and the hierarchical structures formed by the queries. Specifically, to exploit hierarchical structure, we propose a reconciliation learning module. It leverages deep learning model to learn the reconciliation relation between the hierarchical series in the latent space automatically, and forces the coherence constraints through a distill reconciliation loss. To exploit semantic information, we propose a semantic representation module and generate semantic aware series embeddings for queries. Extensive experiments are conducted to confirm the effectiveness of the proposed method.
710,673
Title: Multi-Faceted Hierarchical Multi-Task Learning for Recommender Systems Abstract: ABSTRACTThere have been many studies on improving the efficiency of shared learning in Multi-Task Learning (MTL). Previous works focused on the "micro" sharing perspective for a small number of tasks, while in Recommender Systems (RS) and many other AI applications, we often need to model a large number of tasks. For example, when using MTL to model various user behaviors in RS, if we differentiate new users and new items from old ones, the number of tasks will increase exponentially with multidimensional relations. This work proposes a Multi-Faceted Hierarchical MTL model (MFH) that exploits the multidimensional task relations in large scale MTLs with a nested hierarchical tree structure. MFH maximizes the shared learning through multi-facets of sharing and improves the performance with heterogeneous task tower design. For the first time, MFH addresses the "macro" perspective of shared learning and defines a "switcher" structure to conceptualize the structures of macro shared learning. We evaluate MFH and SOTA models in a large industry video platform of 10 billion samples and hundreds of millions of monthly active users. Results show that MFH outperforms SOTA MTL models significantly in both offline and online evaluations across all user groups, especially remarkable for new users with an online increase of 9.1% in app time per user and 1.85% in next-day retention rate. MFH currently has been deployed in WeSee, Tencent News, QQ Little World and Tencent Video, several products of Tencent. MFH is especially beneficial to the cold-start problems in RS where new users and new items often suffer from a "local overfitting" phenomenon that we first formalize in this paper.
710,674
Title: Knowledge Enhanced Multi-Interest Network for the Generation of Recommendation Candidates Abstract: ABSTRACTCandidate generation task requires that candidates related to user interests need to be extracted in realtime. Previous works usually transform a user's behavior sequence to a unified embedding, which can not reflect the user's multiple interests. Some recent works like Comirec and Octopus use multi-channel structures to capture users' diverse interests. They cluster users' historical behaviors into several groups, claiming that one group represents one interest. However, these methods have some limitations. First, an item may correspond to multiple interests of users, thereby simply allocating it to just one interest group will make the modeling of users' interests coarse-grained and inaccurate. Second, explaining user interests at the level of items is rather vague and not convincing. In this paper, we propose a Knowledge Enhanced Multi-Interest Network: KEMI, which exploits knowledge graphs to help learn users' diverse interest representations via heterogeneous graph neural networks (HGNNs) and a novel dual memory network. Specifically, we use HGNNs to capture the semantic representation of knowledge entities and a novel dual memory network to learn a user's diverse interests from his behavior sequence. Through memory slots of the user memory network and the item memory network, we can learn multiple interests for each user and each item. Meanwhile, by binding the entities to the channels of memory networks, we enable it to be explained from the perspective of the knowledge graph, which enhances the interpretability and understanding of user interests. We conduct extensive experiments on two industrial and publicly available datasets. Experimental results demonstrate that our model achieves significant improvements over state-of-the-art baseline models.
710,675
Title: PlatoGL: Effective and Scalable Deep Graph Learning System for Graph-enhanced Real-Time Recommendation Abstract: ABSTRACTRecently, graph neural network (GNN) approaches have received huge interests in recommendation tasks due to their ability of learning more effective user and item representations. However, existing GNN-based recommendation models cannot support real-time recommendation where the model keeps its freshness by continuously training the streaming data that users produced, leading to negative impact on recommendation performance. To fully support graph-enhanced large-scale recommendation in real-time scenarios, a deep graph learning system is required to dynamically store the streaming data as a graph structure and enable the development of any GNN model incorporated with the capabilities of real-time training and online inference. However, such requirements rule out existing deep graph learning solutions. In this paper, we propose a new deep graph learning system called PlatoGL, where (1) an effective block-based graph storage is designed with non-trivial insertion/deletion mechanism for updating the graph topology in-milliseconds, (2) a non-trivial multi-blocks neighbour sampling method is proposed for efficient graph query, and (3) a cache technique is exploited to improve the storage stability. We have deployed PlatoGL in Wechat, and leveraged its capability in various content recommendation scenarios including live-streaming, article and micro-video. Comprehensive experiments on both deployment performance and benchmark performance~(w.r.t. its key features) demonstrate its effectiveness and scalability. One real-time GNN-based model, developed with PlatoGL, now serves the major online traffic in WeChat live-streaming recommendation scenario.
710,676
Title: Predicting Multi-level Socioeconomic Indicators from Structural Urban Imagery Abstract: ABSTRACTUnderstanding economic development and designing government policies requires accurate and timely measurements of socioeconomic activities. In this paper, we show how to leverage city structural information and urban imagery like satellite images and street view images to accurately predict multi-level socioeconomic indicators. Our framework consists of four steps. First, we extract structural information from cities by transforming real-world street networks into city graphs (GeoStruct). Second, we design a contrastive learning-based model to refine urban image features by looking at geographic similarity between images, with images that are geographically close together having similar features (GeoCLR). Third, we propose using street segments as containers to adaptively fuse the features of multi-view urban images, including satellite images and street view images (GeoFuse). Finally, given the city graph with a street segment as a node and a neighborhood area as a subgraph, we jointly model street- and neighborhood-level socioeconomic indicator predictions as node and subgraph classification tasks. The novelty of our method is that we introduce city structure to organize multi-view urban images and model the relationships between socioeconomic indicators at different levels. We evaluate our framework on the basis of real-world datasets collected in multiple cities. Our proposed framework improves performance by over 10% when compared to state-of-the-art baselines in terms of prediction accuracy and recall.
710,677
Title: Cognitive Diagnosis Focusing on Knowledge Concepts Abstract: ABSTRACTCognitive diagnosis is a crucial task in the field of educational measurement and psychology, which aims to diagnose the strengths and weaknesses of participants. Existing cognitive diagnosis methods only consider which of knowledge concepts are involved in the knowledge components of exercises, but ignore the fact that different knowledge concepts have different effects on practice scores in actual learning situations. Therefore, researchers need to reshape the learning scene by combining the multi-factor relationships between knowledge components. In this paper, in order to more comprehensively simulate the interaction between students and exercises, we developed a neural network-based CDMFKC model for cognitive diagnosis. Our method not only captures the nonlinear interaction between exercise characteristics, student performance, and their mastery of each knowledge concept, but also further considers the impact of knowledge concepts by designing the difficulty and discrimination of knowledge concepts, and uses multiple neural layers to model their interaction so as to obtain accurate and interpretable diagnostic results. In addition, we propose an improved CDMFKC model with guessing parameter and slipping parameter designed by knowledge concept proficiency and student proficiency vectors. We validate the performance of these two diagnostic models on six real datasets. The experimental results show that the two models have better effects in the aspects of accuracy, rationality and interpretability.
710,678
Title: Query Rewriting in TaoBao Search Abstract: ABSTRACTIn e-commerce search engines, query rewriting (QR) is a crucial technique that improves shopping experience by reducing the vocabulary gap between user queries and product catalog. Recent works have mainly adopted the generative paradigm. However, they hardly ensure high-quality generated rewrites and do not consider personalization, which leads to degraded search relevance. In this work, we present Contrastive Learning Enhanced Query Rewriting (CLE-QR), the solution used in Taobao product search. It uses a novel contrastive learning enhanced architecture based on "query retrieval-semantic relevance ranking-online ranking". It finds the rewrites from hundreds of millions of historical queries while considering relevance and personalization. Specifically, we first alleviate the representation degeneration problem during the query retrieval stage by using an unsupervised contrastive loss, and then further propose an interaction-aware matching method to find the beneficial and incremental candidates, thus improving the quality and relevance of candidate queries. We then present a relevance-oriented contrastive pre-training paradigm on the noisy user feedback data to improve semantic ranking performance. Finally, we rank these candidates online with the user profile to model personalization for the retrieval of more relevant products. We evaluate CLE-QR on Taobao Product Search, one of the largest e-commerce platforms in China. Significant metrics gains are observed in online A/B tests. CLE-QR has been deployed to our large-scale commercial retrieval system and serviced hundreds of millions of users since December 2021. We also introduce its online deployment scheme, and share practical lessons and optimization tricks of our lexical match system.
710,679
Title: An Adaptive Framework for Confidence-constraint Rule Set Learning Algorithm in Large Dataset Abstract: ABSTRACTDecision rules have been successfully used in various classification applications because of their interpretability and efficiency. In many real-world scenarios, especially in industrial applications, it is necessary to generate rule sets under certain constraints, such as confidence constraints. However, most previous rule mining methods only emphasize the accuracy of the rule set but take no consideration of these constraints. In this paper, we propose a Confidence-constraint Rule Set Learning (CRSL) framework consisting of three main components, i.e. rule miner, rule ranker, and rule subset selector. Our method not only considers the trade-off between confidence and coverage of the rule set but also considers the trade-off between interpretability and performance. Experiments on benchmark data and large-scale industrial data demonstrate that the proposed method is able to achieve better performance (6.7% and 8.8% improvements) and competitive interpretability when compared with other rule set learning methods.
710,680
Title: PAVE: Lazy-MDP based Ensemble to Improve Recall of Product Attribute Extraction Models Abstract: ABSTRACTE-commerce stores face the challenge of missing and inconsistent attribute values in the product detail pages and have to impute them on behalf of their vendors. Traditional approaches formulate the problem of attribute extraction(AE) from product profiles as natural language tasks such as information extraction or text classification. Such models typically operate at high precision but may yield low recall especially on attributes with an open vocabulary due to 1) missing or incorrect information in product profiles, 2) generalization errors due to lack of contextual understanding, and 3) confidence thresholding to operate at high precision. In this work, we present PAVE: Product Attribute Value Ensemble, a novel reinforcement learning model that usesLazy-MDP formalism to solve for low recall by aggregating information from a sequence of product neighbors. We train a policy network usingProximal Policy Optimization that learns to choose the correct value from the sequence. We observe consistent improvement in recall across all open attributes compared to traditionalAE models with an average lift of 10.3% with no drop in precision. Our method surpasses simple aggregation methods like nearest neighbor, majority vote and binary classifier ensembles and even outperformsAE models for closed attributes. Our approach is scalable, robust to noisy product neighbors and generalizes well on unseen attributes.
710,681
Title: RaDaR: A Real-Word Dataset for AI powered Run-time Detection of Cyber-Attacks Abstract: ABSTRACTArtificial Intelligence techniques on malware run-time behavior have emerged as a promising tool in the arms race against sophisticated and stealthy cyber-attacks. While data of malware run-time features are critical for research and benchmark comparisons, unfortunately, there is a dearth of real-world datasets due to multiple challenges to their collection. The evasive nature of malware, its dependence on connected real-world conditions to execute, and its potential repercussions pose significant challenges for executing malware in laboratory settings. Consequently, prior open datasets rely on isolated virtual sandboxes to run malware, resulting in data that is not representative of malware behavior in the wild. This paper presents RaDaR, an open real-world dataset for run-time behavioral analysis of Windows malware. RaDaR is collected by executing malware on a real-world testbed with Internet connectivity and in a timely manner, thus providing a close-to-real-world representation of malware behavior. To enable an unbiased comparison of different solutions and foster multiple verticals in malware research, RaDaR provides a multi-perspective data collection and labeling of malware activity. The multi-perspective collection provides a comprehensive view of malware activity across the network, operating system (OS), and hardware. On the other hand, the multi-perspective labeling provides four independent perspectives to analyze the same malware, including its methodology, objective, capabilities, and the information it exfiltrates. To date, RaDaR includes 7 million network packets, 11.3 million OS system call traces, and 3.3 million hardware events of 10,434 malware samples having different methodologies (3 classes) and objectives (9 classes), spread across 30 well-known malware families.
710,682
Title: Bridging Self-Attention and Time Series Decomposition for Periodic Forecasting Abstract: ABSTRACTIn this paper, we study how to capture explicit periodicity to boost the accuracy of deep models in univariate time series forecasting. Recent advanced deep learning models such as recurrent neural networks (RNNs) and transformers have reached new heights in terms of modeling sequential data, such as natural languages, due to their powerful expressiveness. However, real-world time series are often more periodic than general sequential data, while recent studies confirm that standard neural networks are not capable of capturing the periodicity sufficiently because they have no modules that can represent periodicity explicitly. In this paper, we alleviate this challenge by bridging the self-attention network with time series decomposition and propose a novel framework called DeepFS. DeepFS equips Deep models with F ourier S eries to preserve the periodicity of time series. Specifically, our model first uses self-attention to encode temporal patterns, from which to predict the periodic and non-periodic components for reconstructing the forecast outputs. The Fourier series is injected as an inductive bias in the periodic component. Capturing periodicity not only boosts the forecasting accuracy but also offers interpretable insights for real-world time series. Extensive empirical analyses on both synthetic and real-world datasets demonstrate the effectiveness of DeepFS. Studies about why and when DeepFS works provide further understanding of our model.
710,683
Title: Incorporating Fairness in Large-scale Evacuation Planning Abstract: ABSTRACTEvacuation planning is an essential part of disaster management where the goal is to relocate people in a safe and orderly manner. Existing research has shown that such problems are hard to approximate and current methods are difficult to scale to real-life applications. We introduce a notion of fairness and two related objectives while studying evacuation planning, namely: minimizing maximum inconvenience and minimizing average inconvenience. We show that both problems are not just NP-hard to solve exactly, but in fact are NP-hard to approximate. On the positive side, we present a heuristic optimization method MIP-LNS, based on the well-known Large Neighborhood Search framework, that can find good approximate solutions in reasonable amount of time. We also consider a multi-objective problem where the goal is to minimize both objectives and solve it using MIP-LNS. We use real-world road network and population data from Harris County in Houston, Texas (a region that needed large-scale evacuations in the past), and apply MIP-LNS to calculate evacuation plans for the area. We compare the quality of the plans in terms of evacuation efficiency and fairness. We find that the solutions to the multi-objective problem are superior in both of these aspects. We also perform statistical tests to show that the solutions are significantly different.
710,684
Title: DuIVRS: A Telephonic Interactive Voice Response System for Large-Scale POI Attribute Acquisition at Baidu Maps Abstract: ABSTRACTThe task of POI attribute acquisition, which aims at completing missing attributes (e.g., POI name, address, status, phone, and open/close time) for a point of interest (POI) or updating existing attribute values of a POI, plays an essential role in enabling users to entertain location-based services using commercial map applications, such as Baidu Maps. Existing solutions have adopted street views or web documents to acquire POI attributes, which have a major limitation in applying for large-scale production due to the labor-intensive and time-consuming nature of collecting data, error accumulation in processing textual/visual data in unstructured or free format, and necessitating post-processing steps with manual efforts. In this paper, we present our efforts and findings from a 3-year longitudinal study on designing and implementing DuIVRS, which is an alternative, fully automatic, and production-proven solution for large-scale POI attribute acquisition via completely machine-directed dialogues. Specifically, DuIVRS is designed to proactively acquire POI attributes via a telephonic interactive voice response system, whose tasks are to generate machine-initiative directed dialogues, make scripted telephone calls to businesses, and interact with people who answered the phone to achieve predefined goals through multi-turn dialogues. DuIVRS has already been deployed in production at Baidu Maps since December 2018, which greatly improves productivity and reduces production cost of POI attribute acquisition. As of December 31, 2021, DuIVRS has made 140 million calls and 42 million POI attribute updates within a 3-year period, which represents an approximately 3-year workload for a high-performance team of 1,000 call center workers. This demonstrates that DuIVRS is an industrial-grade and robust solution for cost-effective, large-scale acquisition of POI attributes.
710,685
Title: BLUTune: Query-informed Multi-stage IBM Db2 Tuning via ML Abstract: ABSTRACTModern data systems such as IBM Db2 have hundreds of system configuration parameters, ''knobs", which heavily influence the performance of business queries. Manual configuration, ''tuning," by experts is painstaking and time consuming. We propose a query informed tuning system called BLUTune which uses machine learning (ML)-deep reinforcement learning based on advantage actor critic neural networks-to tune configurations within defined resource constraints. We translate high-dimensional query execution plans (QEPs) into a low-dimensional embedding space (QEP2Vec) for input into the ML models. To scale to complex and large workloads, we bootstrap the training process through transfer learning. We first train our model based on the estimated cost of queries; we then fine-tune it based on actual query execution times. We demonstrate by an experimental study over various synthetic and real-world workloads BLUTune's efficiency and effectiveness.
710,686
Title: PROPN: Personalized Probabilistic Strategic Parameter Optimization in Recommendations Abstract: ABSTRACTReal-world recommender systems usually consist of two phases. Predictive models in Phase I provide accurate predictions of users' actions on items, and Phase II is to aggregate the predictions withstrategic parameters to make final recommendations, which aim to meet multiple business goals, such as maximizing users' like rate and average engagement time. Though it is important to generate accurate predictions in Phase I, it is also crucial to optimize the strategic parameters in Phase II. Conventional solutions include manually tunning, Bayesian optimization, contextual multi-armed bandit optimization, etc. However, these methods either produce universal strategic parameters for all the users or focus on a deterministic solution, which leads to an undesirable performance. In this paper, we propose a personalized probabilistic solution for strategic parameter optimization. We first formulate the personalized probabilistic optimizing problem and compare its solution with deterministic and context-free solutions theoretically to show its superiority. We then introduce a novel Personalized pRObabilistic strategic parameter optimizing Policy Network (PROPN) to solve the problem. PROPN follows reinforcement learning architecture where a neural network serves as an agent that dynamically adjusts the distributions of strategic parameters for each user. We evaluate our model under the streaming recommendation setting on two public real-world datasets. The results show that our framework outperforms representative baseline methods.
710,687
Title: Sentaur: Sensor Observable Data Model for Smart Spaces Abstract: ABSTRACTThis paper presents Sentaur, a middleware designed, built, and deployed to support sensor-based smart space analytical applications. Sentaur supports a powerful data model that decouples semantic data (about the application domain) from sensor data (using which the semantic data is derived). By supporting mechanisms to map/translate data, concepts, and queries between the two levels, Sentaur relieves application developers from having to know or reason about either capabilities of sensors or write sensor specific code. This paper describes Sentaur's data model, its translation strategy, and highlights its benefits through real-world case studies.
710,688
Title: UDM: A Unified Deep Matching Framework in Recommender Systems Abstract: ABSTRACTDue to the large-scale users and items, industrial recommender systems usually consist of two stages, the matching stage and the ranking stage. The matching stage is responsible for retrieving a small fraction of relevant items from the large-scale item pool which are further selected by the ranking stage. Most of the existing deep learning-based matching models focus on the problem of modeling user interest representation by using inner product between user representation and item representation to obtain the user-to-item relevance. However, the item-to-item relevance between user interacted item and target item is not considered in the deep matching models which is computationally prohibitive for large-scale applications. In this paper, we propose a unified deep matching framework called UDM for the matching stage to mitigate this issue. UDM can model the user-to-item relevance and item-to-item relevance simultaneously with the help of an interest extraction module and interest interaction module, respectively. Specifically, the interest extraction module is used as the main network to extract users' multiple interests with multiple vectors based on users' behavior sequences, while the interest interaction module is used as an auxiliary network to supervise the learning of the interest extraction module, which can model the interaction between user interacted items and target item. In the experiments conducted on two public datasets and a large-scale industrial dataset, UDM achieves consistent improvements over state-of-the-art models. Moreover, UDM has been deployed in the operational system of Alibaba. Online A/B testing results further reveal the effectiveness of UDM. To the best of our knowledge, UDM is the first deep matching framework which combines the user-to-item relevance modeling and item-to-item relevance modeling in the same model.
710,689
Title: Towards Practical Large Scale Non-Linear Semi-Supervised Learning with Balancing Constraints Abstract: ABSTRACTSemi-Supervised Support Vector Machine (S3VM) is one of the most popular methods for semi-supervised learning, which can make full use of plentiful, easily accessible unlabeled data. Balancing constraint is normally enforced in S3VM (denoted as BCS3VM) to avoid the harmful solution which assigns all or most of the unlabeled examples to one same label. Traditionally, non-linear BCS3VM is solved by sequential minimal optimization algorithm. Recently, a novel incremental learning algorithm (IL-BCS3VM) was proposed to scale up BCS3VM further. However, IL-BCS3VM needs to calculate the inverse of the linear system related to the support matrix, making the algorithm not scalable enough. To make BCS3VM be more practical in large-scale problems, in this paper, we propose a new scalable BCS3VM with accelerated triply stochastic gradients (denoted as TSG-BCS3VM). Specifically, to make the balancing constraint handle different proportions of positive and negative samples among labeled and unlabeled data, we propose a soft balancing constraint for S3VM. To make the algorithm scalable, we generate triply stochastic gradients by sampling labeled and unlabeled samples as well as the random features to update the solutions, where Quasi-Monte Carlo (QMC) sampling is utilized on random features to accelerate TSG-BCS3VM further. Our theoretical analysis shows that the convergence rate is O(1/√T) for both diminishing and constant learning rates where T is the number of iterations, which is much better than previous results thanks to the QMC method. Empirical results on a variety of benchmark datasets show that our algorithm not only has a good generalization performance but also enjoys better scalability than existing BCS3VM algorithms.
710,690
Title: DuMapper: Towards Automatic Verification of Large-Scale POIs with Street Views at Baidu Maps Abstract: ABSTRACTWith the increased popularity of mobile devices, Web mapping services have become an indispensable tool in our daily lives. To provide user-satisfied services, such as location searches, the point of interest (POI) database is the fundamental infrastructure, as it archives multimodal information on billions of geographic locations closely related to people's lives, such as a shop or a bank. Therefore, verifying the correctness of a large-scale POI database is vital. To achieve this goal, many industrial companies adopt volunteered geographic information (VGI) platforms that enable thousands of crowdworkers and expert mappers to verify POIs seamlessly; but to do so, they have to spend millions of dollars every year. To save the tremendous labor costs, we devised DuMapper, an automatic system for large-scale POI verification with the multimodal street-view data at Baidu Maps. This paper presents not only DuMapper I, which imitates the process of POI verification conducted by expert mappers, but also proposes DuMapper II, a highly efficient framework to accelerate POI verification by means of deep multimodal embedding and approximate nearest neighbor (ANN) search. DuMapper II takes the signboard image and the coordinates of a real-world place as input to generate a low-dimensional vector, which can be leveraged by ANN algorithms to conduct a more accurate search through billions of archived POIs in the database for verification within milliseconds. Compared with DuMapper I, experimental results demonstrate that DuMapper II can significantly increase the throughput of POI verification by 50 times. DuMapper has already been deployed in production since June 2018, which dramatically improves the productivity and efficiency of POI verification at Baidu Maps. As of December 31, 2021, it has enacted over 405 million iterations of POI verification within a 3.5-year period, representing an approximate workload of 800 high-performance expert mappers.
710,691
Title: Numerical Feature Representation with Hybrid N-ary Encoding Abstract: ABSTRACTNumerical features (e.g., statistical features) are widely used in recommender systems and online advertising. Existing approaches for numerical feature representation in industry are primarily based on discretization. However, hard-discretization based methods (e.g., Equal Distance Discretization) are deficient in continuity while soft-discretization based methods (e.g., AutoDis) lack discriminability. To emphasize both continuity and discriminability for numerical features, we propose an end-to-end representation learning framework named NaryDis. Specifically, NaryDis first leverages hybrid n-ary encoding as an automatic discretization module to generate hybrid-grained discretization results (multiple encoded sequences). Each position of the encoded sequence is assigned with a positional embedding and an intra-ary attention network is leveraged to aggregate the positional embeddings for obtaining ary-wise representations. Then an inter-ary attention is adopted to assemble these representations, which are further constrained by a self-supervised regularization module. Comprehensive experiments on two public datasets are conducted to show the superiority and compatibility of NaryDis. Besides, we deeply investigate the properties of continuity and discriminability. Moreover, we further verify the effectiveness of NaryDis on a large-scale industrial advertisement dataset.
710,692
Title: A Context-Enhanced Transformer with Abbr-Recover Policy for Chinese Abbreviation Prediction Abstract: ABSTRACTChinese abbreviation prediction is very important for various natural language processing tasks such as query understanding and entity linking, since people tend to use the concise abbreviation rather than the full form (name) to mention an entity. The existing models achieve their predictions through sequence labeling, i.e., the binary classification for each character (token) of the full form. However, they only leverage the semantics of the entity itself, overlooking the label dependencies between the tokens, and the rich information of the entity-related texts. In this paper we proposed a Context-Enhanced Transformer with Abbr-Recover policy, namely CETAR, for Chinese abbreviation prediction. CETAR predicts the abbreviation sequence mainly through an iterative decoding process, of which each round consists of an abbreviation and recovery operation. Our extensive experiments upon both general field and specific domain datasets justify that CETAR outperforms the state-of-the-art baselines including sequence labeling models and sequence generation models. Moreover, we have successfully constructed a Chinese abbreviation dataset from the famous tour website Fliggy, and we also shared it at https://github.com/tolerancecky/abbr-0731. The online A/B test on the Fliggy search system shows that 2.03% of conversion rate improvement has been achieved with the predicted abbreviations.
710,693
Title: Fooling MOSS Detection with Pretrained Language Models Abstract: ABSTRACTAs artificial intelligence (AI) technologies become increasingly powerful and prominent in society, their misuse is a growing concern. In educational settings, AI technologies could be used by students to cheat on assignments and exams. In this paper we explore whether transformers can be used to solve introductory level programming assignments while bypassing commonly used AI tools to detect similarities between pieces of software. We find that a student using GPT-J [60] can complete introductory level programming assignments without triggering suspicion from MOSS [2], a widely used software similarity and plagiarism detection tool. This holds despite the fact that GPT-J was not trained on the problems in question and is not provided with any examples to work from. We further find that the code written by GPT-J is diverse in structure, lacking any particular tells that future plagiarism detection techniques may use to try to identify algorithmically generated code. We conclude with a discussion of the ethical and educational implications of large language models and directions for future research.
710,694
Title: A Relevant and Diverse Retrieval-enhanced Data Augmentation Framework for Sequential Recommendation Abstract: ABSTRACTWithin online platforms, it is critical to capture the semantics of sequential user behaviors for accurately predicting user interests. Recently, significant progress has been made in sequential recommendation with deep learning. However, existing neural sequential recommendation models may not perform well in practice due to the sparsity of the real-world data especially in cold-start scenarios. To tackle this problem, we propose the model ReDA, which stands for Retrieval-enhanced Data Augmentation for modeling sequential user behaviors. The main idea of our approach is to leverage the related information from similar users for generating both relevant and diverse augmentation. First, we train a neural retriever to retrieve the augmentation users according to the se- mantic similarity between user representations, and then conduct two types of data augmentation to generate augmented user representations. Furthermore, these augmented data are incorporated in a contrastive learning framework for learning more capable representations. Extensive experiments conducted on both public and industry datasets demonstrate the superiority of our proposed method over existing state-of-the-art methods, especially when only limited training data is available.
710,695
Title: Debiased Balanced Interleaving at Amazon Search Abstract: ABSTRACTInterleaving is an online evaluation technique that has shown to be orders of magnitude more sensitive than traditional A/B tests. It presents users with a single merged result of the compared rankings and then attributes user actions back to the evaluated rankers. Different interleaving methods in the literature have their advantages and limitations with respect to unbiasedness, sensitivity, preservation of user experience, and implementation and computation complexity. We propose a new interleaving method that utilizes a counterfactual evaluation framework for credit attribution while sticking to the simple ranking merge policy of balanced interleaving, and formally derive an unbiased estimator for comparing rankers with theoretical guarantees. We then confirm the effectiveness of our method with both synthetic and real experiments. We also discuss practical considerations of bringing different interleaving methods from the literature into a large-scale experiment, and show that our method achieves a favorable tradeoff in implementation and computation complexity while preserving statistical power and reliability. We have successfully implemented our method and produced consistent conclusions at the scale of billions of search queries. We report 10 online experiments that apply our method to e-commerce search, and observe a 60x sensitivity gain over A/B tests. We also find high correlations between our proposed estimator and corresponding A/B metrics, which helps interpret interleaving results in the magnitude of A/B measurements.
710,696
Title: Graph Neural Networks Pretraining Through Inherent Supervision for Molecular Property Prediction Abstract: ABSTRACTRecent global events have emphasized the importance of accelerating the drug discovery process. A way to deal with the issue is to use machine learning to increase the rate at which drugs are made available to the public. However, chemical labeled data for real-world applications is extremely scarce making traditional approaches less effective. A fruitful course of action for this challenge is to pretrain a model using related tasks with large enough datasets, with the next step being finetuning it for the desired task. This is challenging as creating these datasets requires labeled data or expert knowledge. To aid in solving this pressing issue, we introduce MISU - Molecular Inherent SUpervision, a unique method for pretraining graph neural networks for molecular property prediction. Our method leapfrogs past the need for labeled data or any expert knowledge by introducing three innovative components that utilize inherent properties of molecular graphs to induce information extraction at different scales, from the local neighborhood of an atom to substructures in the entire molecule. Our empirical results for six chemical-property-prediction tasks show that our method reaches state-of-the-art results compared to numerous baselines.
710,697
Title: Efficient and Effective SPARQL Autocompletion on Very Large Knowledge Graphs Abstract: ABSTRACTWe show how to achieve fast autocompletion for SPARQL queries on very large knowledge graphs. At any position in the body of a SPARQL query, the autocompletion suggests matching subjects, predicates, or objects. The suggestions are context-sensitive and ranked by their relevance to the part of the query already typed. The suggestions can be narrowed down by prefix search on the names and aliases of the desired subject, predicate, or object. All suggestions are themselves obtained via SPARQL queries. For existing SPARQL engines, these queries are impractically slow on large knowledge graphs. We present various algorithmic and engineering improvements of an open-source SPARQL engine such that these queries are executed efficiently. We evaluate a variety of suggestion methods on three large knowledge graphs, including the complete Wikidata. We compare our results with two widely used SPARQL engines, Virtuoso and Blazegraph. Our code, benchmarks, and complete reproducibility materials are available on https://ad.cs.uni-freiburg.de/publications.
710,698
Title: Will This Online Shopping Session Succeed? Predicting Customer's Purchase Intention Using Embeddings Abstract: ABSTRACTCustomers are increasingly using online channels to buy products. For e-commerce companies, this offers new opportunities to tailor the shopping experience to customers' needs. Therefore, it is of great importance for a company to know their customers' intentions while browsing their webpage. A major challenge is the real-time analysis of a customer's intention during browsing sessions. To this end, a representation of the customer's browsing behavior must be retrieved from their live interactions on the webpage. Typically, characteristic behavioral features are extracted manually based on the knowledge of marketing experts. In this paper, we propose a customer embedding representation that is based on the customer's click-events recorded during browsing sessions. Thus, our approach does not use manually extracted features and is not based on marketing expert domain knowledge, which makes it transferable to different webpages and different online markets. We demonstrate our approach using three different e-commerce datasets to successfully predict whether a customer is going to purchase a specific product. For the prediction, we utilize the customer embedding representations as input for different machine learning models. We compare our approach with existing state-of-the-art approaches for real-time purchase prediction and show that our proposed customer representation with an LSTM predictor outperforms the state-of-the-art approach on all three datasets. Additionally, the creation process of our customers' representation is on average 235 times faster than the creation process of the baseline.
710,699
Title: How Hybrid Work Will Make Work More Intelligent Abstract: ABSTRACTWe are in the middle of the most significant change to work practices in generations. For hundreds of years, physical space was the most important technology people used to get things done. The coming Hybrid Work Era, however, will be shaped by digital technology. The recent rapid shift to remote work accelerated the digital transformation already underway at many organizations, and new types of work-related data are now being generated at an unprecedented rate. For example, the average Microsoft Teams user spends 252% more time in the application now than they did in February 2020. During the early stages of the pandemic, we saw the direct impact of digital technology on work in its ability to help people sustain collaboration across time and space. But looking forward, the new digital knowledge captured in the Hybrid Work Era will allow us to reimagine work at an even more fundamental level. AI systems, for example, can now learn from the conversations people have to support knowledge re-use, and even learn how successful conversations happen to help drive more productive meetings. Historically, AI systems have been hindered in a work context by a lack of data; the development of foundation models is changing that, creating an opportunity to combine general world knowledge with the knowledge and behaviors currently locked up and siloed as we work. The CIKM community can shape the new future of work, but first must address the challenges surrounding workplace knowledge management that arise as we have more data, more sophisticated AI, and more human engagement. In this talk I will give an overview of what research tells us about emerging work practices, and explore how the CIKM community can build on these findings to help create a new – and better – future of work.
710,700
Title: Executable Knowledge Graph for Transparent Machine Learning in Welding Monitoring at Bosch Abstract: ABSTRACTWith the development of Industry 4.0 technology, modern industries such as Bosch's welding monitoring witnessed the rapid widespread of machine learning (ML) based data analytical applications, which in the case of welding monitoring has led to more efficient and accurate welding monitoring quality. However, industrial ML is affected by the low transparency of ML towards non-ML experts needs. The lack of understanding by domain experts of ML methods hampers the application of ML methods in industry and the reuse of developed ML pipelines, as ML methods are often developed in an ad hoc manner for specific problems. To address these challenges, we propose the concept and a system of executable Knowledge Graph (KG), which formally encode ML knowledge and solutions in KGs, which serve as common language between ML experts and non-ML experts, thus facilitate their communication and increase the transparency of ML methods. We evaluated our system extensively with an industrial use case at Bosch, showing promising results.
710,701
Title: On the Challenges of Podcast Search at Spotify Abstract: ABSTRACTOnline music streaming is enjoying ever-growing popularity in the last decades, enabled by the abundance of music content in digital format and online streaming services. In the recent years, podcasts, as a talk-focused media format, have witnessed a rapid growth among listeners. Podcasts come in many forms and sizes. They range from 20-minute daily meditation sessions, weekly recaps of global news, interviews with celebrities, and hosts bantering with each other for hours. More and more streaming services are now expanding their catalogs to support both music and podcasts on the same platform, such as Amazon Music, Pandora and Spotify. This setup requires an effective aggregated search system to assemble information from heterogeneous information sources and content types (e.g., artist profiles, playlists, podcast shows, etc.) into one result interface in order to support diverse information needs.
710,702
Title: Geographical Address Models in the Indian e-Commerce Abstract: ABSTRACTUnambiguous customer addresses are important for the e-Commerce companies in timely and accurate delivery of shipments. In many developing countries a prescribed structure is not usually followed in practice. It is observed that the customer addresses contain additional text such as instructions to the delivery team. Further not every address has an associated geolocation information in many countries. Thus address understanding, address classification, and clustering of similar but noisy addresses are critical. In addition to last mile delivery solutions, the models help reduce buyer fraud and understanding returns. In view of noisy nature of customer addresses and non-availability of associated geolocation, the above problems are effectively solved with the help of NLP. The proposed talk traces the challenges in the Indian addresses in the context of e-commerce, solution approaches, and their extensions during the last 8 years. The talk is based on the author's own experience, his publications as well as developments in this topic across the industry over these years.
710,703
Title: Leveraging Automated Search Relevance Evaluation to Improve System Deployment: A Case Study in Healthcare Abstract: ABSTRACTOver the last year, a digital initiative has been focused on reengineering the search engine for kp.org, a health web portal serving over 12 million members. However, traditional software testing techniques that rely on limited use cases and consistent behavior are neither comprehensive nor specific for capturing complex user search behaviors. To support system deployment, we utilize information retrieval (IR) technologies to monitor search performance, identify areas of improvement and suggest actionable items. In this case study we share industrial experience on building an IR evaluation pipeline and its usage to inform deployment and improve system development. The work emphasizes domain specific challenges, best practices and lessons learned during system deployment in a healthcare setting. It features the ability of IR techniques to strengthen collaboration between data scientists, software engineers and product managers in making data-driven decisions.
710,704
Title: Fifty Shades of Pink: Understanding Color in e-commerce using Knowledge Graphs Abstract: ABSTRACTThe color of the products is one of the most prevalent aspects in many e-commerce domains, and it is one of the decisive purchasing factors. Besides having thousands of color variations and shades, many brands continuously develop proprietary colors and color names to attract more customers. This often leads to color ambiguity (textual and visual), and vocabulary mismatch between buyers and sellers. Therefore, it is crucial for any e-commerce search engine to correctly identify the buyer's color intent and match it to the corresponding product listings. To address this challenge, in this work, we introduce a color query expansion approach using color Knowledge Graphs. We use Knowledge Graphs to unambiguously identify all the colors based on their properties, and the relationships to other colors, which allows us to perform semantic query expansion. Similar expansion concepts could be applied to domains outside of color.
710,705
Title: Building Next Best Action Engines for B2C and B2B Use Cases Abstract: ABSTRACTTraditional machine learning methods used in marketing and digital commerce applications, including propensity scoring and many recommendation algorithms, are usually focused on improving short-term outcomes such as click-through rates. In many environments, however, long-term customer engagement can be more important than immediate outcomes. In this paper, we describe several real-world case studies on building personalization engines that address this problem using reinforcement learning (RL) methods. We also discuss the design patterns used to create such solutions.
710,706
Title: Simulating Complex Problems Inside a Database Abstract: ABSTRACTThe standard way to store and interact with the large amount of data that are central to the functioning of any modern business is through the use of a relational Knowledge Graph Management System (KGMS). In this paper we show how the relational model can be successfully exploited to model complex analytic scenarios while enjoying the same characteristics of clarity and flexibility as when modeling the data themselves. Using the Rel language, we simulate the daily schedule of an airline company as an agentbased system, and we will show how modeling this system through a set of relationships and logical rules will let us focus directly on the inherent complexity of our model, taking away most of the incidental effort in actually implementing our simulation.
710,707
Title: From Product Searches to Conversational Agents for E-Commerce Abstract: ABSTRACTAs consumers' demand for online shopping substantially increased in the last few years, e-commerce companies are still far from providing a high-quality user experience that may compete with in-store experiences. On the one hand, matching search queries with highly relevant products for discovery and browsing is still a challenge within existing search technologies. Available e-commerce solutions hardly provide tools to optimize product search relevance and fail to integrate user behavior signals into the search optimization pipeline. On the other hand, accessing the rich and complex information concealed in an e-commerce catalog through a search bar has not evolved far since its initial adoption. In this talk, we illustrate how the VUI conversational AI platform has been successfully adopted to both improve the user's experience quality with highly relevant search and discovery results and expand the traditional search bar with conversational agents' technology, enriching the user's experience at each stage of the e-commerce product life cycle. We review in depth some of the key deep learning models as part of the query understanding component and discuss the overall conversation architecture as it integrates with an existing e-commerce catalog. We include real-life demonstrations derived from use cases extracted from deployed systems.
710,708
Title: Synerise Monad - Real-Time Multimodal Behavioral Modeling Abstract: ABSTRACTThe growth of time-sensitive heterogeneous data in industry-grade datalakes has recently reached unprecedented momentum. In response to this, we propose Synerise Monad - a prototype of a real-time behavioral modeling platform for event-based data streams. It automates representation learning and model training on massive data sources with arbitrary data structures. With Monad we showcase how to automatically process various data modalities, such as temporal, graph, categorical, decimal, and textual data types, in a time-sensitive way allowing for real-time time feature creation and predictions. Monad's distributed and scalable architecture coupled with efficient award-winning algorithms developed at Synerise - Cleora and EMDE - allows to process real-life datasets composed of billions of events in record time. The Monad ecosystem showcases a viable path towards real-time event-based AutoML.
710,709
Title: Utilizing Contrastive Learning To Address Long Tail Issue in Product Categorization Abstract: ABSTRACTNeural network models trained in a supervised learning way have become dominant. Although high performances can be achieved when training data is ample, the performance on labels with sparse training instances can be poor. This performance drift caused by imbalanced data is named as long tail issue and impacts many NN models used in reality. In this talk, we will firstly review machine learning approaches addressing the long-tail issue. Next, we will report on our effort on applying one recent LT-addressing method on the item categorization (IC) task that aims to classify product description texts into leaf nodes in a category taxonomy tree. In particular, we adopted a new method, which consists of decoupling the entire classification task into (a) learning representations using the K-positive contrastive loss (KCL) and (b) training a classifier on balanced data set, into IC tasks. Using SimCSE to be our self-learning backbone, we demonstrated that the proposed method works on the IC text classification task. In addition, we spotted a shortcoming in the KCL: false negative (FN) instances may harm the representation learning step. After eliminating FN instances, IC performance (measured by macro-F1) has been further improved.
710,710
Title: Intent Disambiguation for Task-oriented Dialogue Systems Abstract: ABSTRACTTask-Oriented Dialogue Systems (TODS) have been widely deployed for domain specific virtual assistants at contact centres to route customers' calls or deliver information needs of the customer in a conversational interaction. TODS employ natural language understanding components in order to map user commands to a set of pre-defined intents. However, Contact Centre users often fail to formulate their complex information needs in a single utterance which leads to formulating ambiguous user commands. This can negatively impact intent classification, and consequently customer satisfaction. To avoid feeding ambiguous user commands to the intent classifier of virtual assistants and help users in formulating their commands, we have implemented a solution that (1) identifies when a user is ambiguous and the virtual assistant should ask a clarification question, (2) disambiguates the user command and provides top-N most likely intents in a form of a clarification question. Our experimental result shows that our proposed intent disambiguation solution has a statistically significant improvement over a popularity based intent disambiguation model and an Intent Ranking Model of the Natural Language Understanding engine for a virtual assistant of a contact centre in terms of intent disambiguation accuracy and Mean Reciprocal Rank.
710,711
Title: Modeling Turn-Based Sequences for Player Tactic Applications in Badminton Matches Abstract: ABSTRACTIn recent years, a growing body of research has started to explore applying artificial intelligence to the sports industry due to the availability of data and the advancement of techniques. Such applications not only play an important role during matches but also have a great influence on the training stage. However, applying deep learning techniques to sports analytics has several technical challenges, which also remain untouched in badminton analytics as there is no public dataset of stroke event records. Therefore, this dissertation intends to explore challenging research questions and application issues that have not been addressed for benefiting both the research and badminton communities. To achieve the objectives, we, for the first time, propose a unified badminton language to describe the process of the shot, which enables us to conduct downstream applications. Specifically, our first task is to measure the win probability of each shot in badminton matches by considering long-term and short-term dependencies. Second, we introduce a framework with two encoder-decoder extractors and a position-aware fusion network to forecast the possible tactics of players, which is still unexplored in turn-based sports. To provide the transparency of these models, we aim to design a post-hoc explainer by computing feature attributions with Shapley values as the third task. In this manner, researchers can investigate model behaviors for advanced improvement, and the badminton community benefits from coaching the players and determining the strategies. This dissertation is supervised by Wen-Chih Peng ([email protected]).
710,712
Title: Identify Relevant Entities Through Text Understanding Abstract: ABSTRACTAn Entity Retrieval system is a fundamental task of Information Retrieval that provides direct answer to an information need of user. Prior work of entity retrieval utilizes either the Knowledge Graph fields or the text relevant to the query via pseudo-relevance feedback to improve the performance. Recently, Knowledge Graph embeddings or other entity representations, which capture the entity information from a knowledge graph are shown to be beneficial for entity retrieval. However, such embeddings are query-agnostic. In this dissertation work, we aim to improve entity retrieval by exploring the pseudo-relevance feedback to generate entity representations that capture query-aware entity information to determine the relevance of entities. We study the effectiveness of pseudo-relevance feedback against Knowledge Graph fields and investigate the efficacy of the Knowledge Graph embeddings for entity retrieval. We aim to understand the importance of utilization of query-aware signals and modeling of such signals with Knowledge Graph embeddings. Our results show that pseudo-relevance feedback is more effective than the Knowledge Graph fields by 30%.
710,713
Title: Best Practices for Top-N Recommendation Evaluation: Candidate Set Sampling and Statistical Inference Techniques Abstract: ABSTRACTTop-N recommendation evaluation experiments are complex, with many decisions needed. These decisions are often made inconsistently, and we don't have clear best practices for many of them. The goal of this project, is to identify, substantiate, and document best practices to improve evaluations.
710,714
Title: Causal Relationship over Knowledge Graphs Abstract: ABSTRACTCausality has been discussed for centuries, and the theory of causal inference over tabular data has been broadly studied and utilized in multiple disciplines. However, only a few works attempt to infer the causality while exploiting the meaning of the data represented in a data structure like knowledge graph. These works offer a glance at the possibilities of causal inference over knowledge graphs, but do not yet consider the metadata, e.g., cardinalities, class subsumption and overlap, and integrity constraints. We propose CareKG, a new formalism to express causal relationships among concepts, i.e., classes and relations, and enable causal queries over knowledge graphs using semantics of metadata. We empirically evaluate the expressiveness of CareKG in a synthetic knowledge graph concerning cardinalities, class subsumption and overlap, integrity constraints. Our initial results indicate that CareKG can represent and measure causal relations with some semantics which are uncovered by state-of-the-art approaches.
710,715
Title: C-Cast: A Real-Time Forecasting Model for a Controlled Sequence Abstract: ABSTRACTPredictive control is an advanced control method that is used successfully in industrial control applications. One of the most fundamental demands for predictive control is the accurate forecasting of a ''controlled sequence" using exogenous sequences which consist of multiple attributes (manipulated variables and operation signals). Given a controlled sequence and exogenous sequences, how can we effectively forecast the future behavior of a controlled sequence? In this paper, we present C-Cast, an efficient and effective method for forecasting a time-evolving controlled sequence with exogenous sequences. Our proposed method has the following properties: (a) Adaptive: it captures important time-evolving patterns and operation shift in a time-evolving controlled sequence (b) Effective: it performs accurate forecasting. (c) Practical: it enables real-time controlled sequence forecasting fast enough to satisfy the limitation required for predictive control. Extensive experiments on a real dataset demonstrate that C-Cast consistently outperforms the best existing state-of-the-art methods as regards accuracy, and the execution speed is sufficiently fast.
710,716
Title: Rank-Aware Gain-Based Evaluation of Extractive Summarization Abstract: ABSTRACTROUGE has long been a popular metric for evaluating text summarization tasks as it eliminates time-consuming and costly human evaluations. However, ROUGE is not a fair evaluation metric for extractive summarization task as it is entirely based on lexical overlap. Additionally, ROUGE ignores the quality of the ranker for extractive summarization which performs the actual sentence/phrase extraction job. The main focus of the thesis is to design a nCG (normalized cumulative gain)-based evaluation metric for extractive summarization that is both rank-aware and semantic-aware (called Sem-nCG). One fundamental contribution of the work is that it demonstrates how we can generate more reliable semantic-aware ground truths for evaluating extractive summarization tasks without any additional human intervention. To the best of our knowledge, this work is the first of its kind. Preliminary experimental results demonstrate that the new Sem-nCG metric is indeed semantic-aware and also exhibits higher correlation with human judgement for single document summarization when single reference is considered.
710,717
Title: Sequence-Driven Analytics and Prediction Abstract: ABSTRACTExisting algorithms for sequential patterns do not address data heterogeneity. These algorithms can mine unusual patterns in the database and use them for the prediction. Traditional mining methods ignore pattern usefulness concerning heterogeneity and uncertainty. Some patterns may often exist in the database but may not add to its overall usefulness. Due to the changing database, this field can undertake more effective mining and prediction. Data security has become a priority with more data and user privacy problems. Laws and regulations help develop a safe data society, but performing pattern analysis without sharing data presents a new problem. Now, legally mandated data protection rules create issues with new data regulations and laws that require more attention. First, we have investigated conventional sequential mining methods that can improve effectiveness and efficiency in a heterogeneous environment. Second, we investigate learning and evolution methods for privacy protection. In the following, we briefly describe the insights we gained by publishing the defined research.
710,718
Title: Tutorial on Deep Learning Interpretation: A Data Perspective Abstract: ABSTRACTDeep learning models have achieved exceptional predictive performance in a wide variety of tasks, ranging from computer vision, natural language processing, to graph mining. Many businesses and organizations across diverse domains are now building large-scale applications based on deep learning. However, there are growing concerns, regarding the fairness, security, and trustworthiness of these models, largely due to the opaque nature of their decision processes. Recently, there has been an increasing interest in explainable deep learning that aims to reduce the opacity of a model by explaining its behavior, its predictions, or both, thus building trust between human and complex deep learning models. A collection of explanation methods have been proposed in recent years that address the problem of low explainability and opaqueness of models. In this tutorial, we introduce recent explanation methods from a data perspective, targeting models that process image data, text data, and graph data, respectively. We will compare their strengths and limitations, and offer real-world applications.
710,719
Title: Mining of Real-world Hypergraphs: Patterns, Tools, and Generators Abstract: ABSTRACTGroup interactions are prevalent in various complex systems (e.g., collaborations of researchers and group discussions on online Q&A sites), and they are commonly modeled as hypergraphs. Hyperedges, which compose a hypergraph, are non-empty subsets of any number of nodes, and thus each hyperedge naturally represents a group interaction among entities. The higher-order nature of hypergraphs brings about unique structural properties that have not been considered in ordinary pairwise graphs. In this tutorial, we offer a comprehensive overview of a new research topic called hypergraph mining. We first present recently revealed structural properties of real-world hypergraphs, including (a) static and dynamic patterns, (b) global and local patterns, and (c) connectivity and overlapping patterns. Together with the patterns, we describe advanced data mining tools used for their discovery. Lastly, we introduce simple yet realistic hypergraph generative models that provide an explanation of the structural properties.
710,720
Title: Graph-based Management and Mining of Blockchain Data Abstract: ABSTRACTThe mainstream adoption of blockchains led to the preparation of many decentralized applications and web platforms, including Web 3.0, a peer-to-peer internet with no single authority. The data stored in blockchain can be considered as big data -- massive-volume, dynamic, and heterogeneous. Due to highly connected structure, graph-based modeling is an optimal tool to analyze the data stored in blockchains. Recently, several research works performed graph analysis on the publicly available blockchain data to reveal insights into its business transactions and for critical downstream tasks, e.g., cryptocurrency price prediction, phishing scams and counterfeit token detection. In this tutorial, we discuss relevant literature on blockchain data structures, storage, categories, data extraction and graphs construction, graph mining, topological data analysis, and machine learning methods used, target applications, and the new insights revealed by them, aiming towards providing a clear view of unified graph-data models for UTXO and account-based blockchains. We also emphasize future research directions.
710,721
Title: Self-Supervised Learning for Recommendation Abstract: ABSTRACTRecommender systems are playing an increasingly critical role to alleviate information overload and satisfy users' information seeking requirements in a wide spectrum of online platforms. However, the ubiquity of data sparsity and noise notably limits the representation capacity of existing recommender systems to learn high-quality user (item) embeddings. Inspired by recent advances of self-supervised learning (SSL) techniques, SSL-based representation learning models benefit a variety of recommendation domains. Such methods have achieved new levels of performance while reducing the dependence on observed supervision labels in diverse recommendation tasks. In this tutorial, we aim to provide a systemic review of state-of-the-art SSL-based recommender systems. To be specific, we summarize and categorize existing work of SSL-based recommender systems in terms of recommendation scenarios. For each type of recommendation task, the corresponding challenges and methods will be presented in a comprehensive way. Finally, some future directions and open questions will be raised to inspire more investigation on this important research line.
710,722