Aarrghh xD

#14
by fconti - opened

Thank you for organizing the competition! it was one of my first competitions and it was a lot of fun. I liked the speechbrain library and the dataset!

I gnawed a bit because one of my solutions was much better than the selected ones, but that's machine learning!

image.png

I'm curious, what approach did you take? Most of the stuff I tried didn't outperform the baseline and my best performing model is a slight modification of the baseline only.

I fine-tuned the embedding model in the base-line without the classifier. I employed a triplet loss approach with online hard batch mining. This allowed me to utilize all the sequences available in the training set!

I am going to share my code soon!

Luck is always an important factor in competition. Me too, after trying some different approaches, I was back to finetuning the baseline model. Would love to learn about your triplet loss approach @fconti !

Congrats to all three of you @fconti @srviksit @dhoa !
Hopefully in next challenge we will be able to make public set larger to minimize surprise in the end, although I must agree that luck is a big thing in these competitions

Thanks for your careful support during this competition @gorinars !!

Sign up or log in to comment