File size: 4,083 Bytes
c633a0a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 |
[
{
"id": 0,
"page": 2,
"bounding_box": [
113.38600158691406,
419.5950012207031,
505.23199462890625,
497.9010009765625
],
"latex_content": "\\begin{table}[htpb]\n\\begin{center}\n\\begin{footnotesize}\n\\newcommand{\\tabincell}[2]{\\begin{tabular}{@{}#1@{}}#2\\end{tabular}}\n\\begin{tabular}{c|c|c|c|c}\\hline\nApproach & Communication & Assumptions & Min signal strength & Strength type \\\\\\hline\nLasso &0 &\\tabincell{c}{Mutual Incoherence \\\\Sparse Eigenvalue} & $\\sqrt{\\frac{\\log p}{n}}$ & Element-wise\\\\\\hline\nGroup lasso &$\\Ocal(np)$ & \\tabincell{c}{Mutual Incoherence \\\\Sparse Eigenvalue} &$\\sqrt{\\frac{1}{n}\\rbr{1 + \\frac{\\log p}{m}}}$ &Row-wise \\\\\\hline \nDSML & $\\Ocal(p)$ &\\tabincell{c}{Generalized Coherence \\\\ Restricted Eigenvalue} &$\\sqrt{\\frac{1}{n}\\rbr{1 + \\frac{\\log p}{m}}} + \\frac{|S|\\log p}{n}$ &Row-wise \\\\\\hline\n\\end{tabular}\n\\small \\caption{\\small Lower bound on coefficients required to\n ensure support recovery with $p$ variables, $m$ tasks, $n$\n samples per task and a true support of size $|S|$.}\n\\label{tab:comparison_sparsistency}\n\\end{footnotesize}\n\\end{center}\n\\end{table}",
"extracted_content": [
[
"Approach",
"Communication",
"Assumptions",
"Min signal strength",
"Strength type"
],
[
"Lasso",
"0",
"Mutual Incoherence\nSparse Eigenvalue",
"q\nlog p\nn",
"Element-wise"
],
[
"Group lasso",
"O(np)",
"Mutual Incoherence\nSparse Eigenvalue",
"q\n1 \u00001 + log p\u0001\nn m",
"Row-wise"
],
[
"DSML",
"O(p)",
"Generalized Coherence\nRestricted Eigenvalue",
"q\nn1 \u00001 + lo mg p\u0001+ |S| nlog p",
"Row-wise"
]
],
"similarity_score": 0.6642335766423357,
"table_image": "images/1510.00633v1/table_0.png",
"page_image": "pages/1510.00633v1/page_2.png"
},
{
"id": 1,
"page": 3,
"bounding_box": [
116.75,
95.239990234375,
495.25,
162.74700927734375
],
"latex_content": "\\begin{table}[htpb]\n\\begin{footnotesize}\n\\newcommand{\\tabincell}[2]{\\begin{tabular}{@{}#1@{}}#2\\end{tabular}}\n\\begin{center}\n\\begin{tabular}{c|c|c|c}\\hline\nApproach & Assumptions & $\\ell_1/\\ell_2$ estimation error & Prediction error \\\\\\hline\nLasso &Restricted Eigenvalue & $ \\sqrt{\\frac{|S|^2 \\log p}{n}}$ & $\\frac{|S| \\log p}{n}$ \\\\\\hline\nGroup lasso & Restricted Eigenvalue & $\\frac{|S|}{\\sqrt{n}} \\sqrt{1 + \\frac{\\log p}{m}}$ &$\\frac{|S|}{n} \\rbr{1 + \\frac{\\log p}{m}}$ \\\\\\hline \nDSML &\\tabincell{c}{Generalized Coherence \\\\ Restricted Eigenvalue} & $\\frac{|S|}{\\sqrt{n}} \\sqrt{1 + \\frac{\\log p}{m}} + \\frac{|S|^2 \\log p}{n}$ &$\\frac{|S|}{n} \\rbr{1 + \\frac{\\log p}{m}} + \\frac{|S|^3 (\\log p)^2}{n^2}$ \\\\\\hline\n\\end{tabular}\n\\end{center}\n\\small \\caption{\\small Comparison of parameter estimation errors and\n prediction errors. The DSML guarantees improve over Lasso and have\n the same leading term as the Group lasso as long as $m<n/(|S|^2\\log\n p)$.}\n\\label{tab:comparison_estimation}\n\\end{footnotesize}\n\\end{table}",
"extracted_content": [
[
"Approach",
"Assumptions",
"\u2113 1/\u2113 estimation error\n2",
"Prediction error"
],
[
"Lasso",
"Restricted Eigenvalue",
"q\n|S|2 log p\nn",
"|S| log p\nn"
],
[
"Group lasso",
"Restricted Eigenvalue",
"q\n\u221a|S n| 1 + lo mg p",
"|S|\u00001 + log p\u0001\nn m"
],
[
"DSML",
"Generalized Coherence\nRestricted Eigenvalue",
"\u221a|S n|q 1 + lo mg p + |S|2 log p\nn",
"|S n|\u00001 + lo mg p\u0001+ |S|3( nlo 2g p)2"
]
],
"similarity_score": 0.3971631205673759,
"table_image": "images/1510.00633v1/table_1.png",
"page_image": "pages/1510.00633v1/page_3.png"
}
] |