We introduce a new estimator for the vector of coefficients \$\beta\$ in the
linear model \$y=X\beta+z\$, where \$X\$ has dimensions \$np\$ with \$p\$
possibly larger than \$n\$. SLOPE, short for Sorted L-One Penalized Estimation,
is the solution to \\min\_bınR^p12\Vert y-Xb\Vert
\_\ell\_2^2+łambda\_1b\_(1)+łambda\_2\vert
b\vert\_(2)+\cdots+łambda\_pb\vert\_(p),\ where
\$łambda\_1\gełambda\_2\ge\cdots\gełambda\_p\ge0\$ and \$\vert
b\vert\_(1)\geb\vert\_(2)\ge\cdots\geb\vert\_(p)\$ are the
decreasing absolute values of the entries of \$b\$. This is a convex program and
we demonstrate a solution algorithm whose computational complexity is roughly
comparable to that of classical \$\ell\_1\$ procedures such as the Lasso. Here,
the regularizer is a sorted \$\ell\_1\$ norm, which penalizes the regression
coefficients according to their rank: the higher the rank - that is, stronger
the signal - the larger the penalty. This is similar to the Benjamini and
Hochberg J. Roy. Statist. Soc. Ser. B 57 (1995) 289-300 procedure (BH) which
compares more significant \$p\$-values with more stringent thresholds. One
notable choice of the sequence \$\łambda\_i\\$ is given by the BH critical
values \$łambda\_BH(i)=z(1-iq/2p)\$, where \$qın(0,1)\$ and
\$z(\alpha)\$ is the quantile of a standard normal distribution. SLOPE aims to
provide finite sample guarantees on the selected model; of special interest is
the false discovery rate (FDR), defined as the expected proportion of
irrelevant regressors among all selected predictors. Under orthogonal designs,
SLOPE with \$łambda\_BH\$ provably controls FDR at level \$q\$.
Moreover, it also appears to have appreciable inferential properties under more
general designs \$X\$ while having substantial power, as demonstrated in a series
of experiments running on both simulated and real data.
%0 Journal Article
%1 bogdan2015slope
%A Bogdan, Małgorzata
%A van den Berg, Ewout
%A Sabatti, Chiara
%A Su, Weijie
%A Candès, Emmanuel J.
%D 2015
%J The Annals of Applied Statistics
%K owl slope structured_sparsity
%N 3
%P 1103--1140
%R 10.1214/15-aoas842
%T SLOPE - Adaptive variable selection via convex optimization
%U http://dx.doi.org/10.1214/15-aoas842
%V 9
%X We introduce a new estimator for the vector of coefficients \$\beta\$ in the
linear model \$y=X\beta+z\$, where \$X\$ has dimensions \$np\$ with \$p\$
possibly larger than \$n\$. SLOPE, short for Sorted L-One Penalized Estimation,
is the solution to \\min\_bınR^p12\Vert y-Xb\Vert
\_\ell\_2^2+łambda\_1b\_(1)+łambda\_2\vert
b\vert\_(2)+\cdots+łambda\_pb\vert\_(p),\ where
\$łambda\_1\gełambda\_2\ge\cdots\gełambda\_p\ge0\$ and \$\vert
b\vert\_(1)\geb\vert\_(2)\ge\cdots\geb\vert\_(p)\$ are the
decreasing absolute values of the entries of \$b\$. This is a convex program and
we demonstrate a solution algorithm whose computational complexity is roughly
comparable to that of classical \$\ell\_1\$ procedures such as the Lasso. Here,
the regularizer is a sorted \$\ell\_1\$ norm, which penalizes the regression
coefficients according to their rank: the higher the rank - that is, stronger
the signal - the larger the penalty. This is similar to the Benjamini and
Hochberg J. Roy. Statist. Soc. Ser. B 57 (1995) 289-300 procedure (BH) which
compares more significant \$p\$-values with more stringent thresholds. One
notable choice of the sequence \$\łambda\_i\\$ is given by the BH critical
values \$łambda\_BH(i)=z(1-iq/2p)\$, where \$qın(0,1)\$ and
\$z(\alpha)\$ is the quantile of a standard normal distribution. SLOPE aims to
provide finite sample guarantees on the selected model; of special interest is
the false discovery rate (FDR), defined as the expected proportion of
irrelevant regressors among all selected predictors. Under orthogonal designs,
SLOPE with \$łambda\_BH\$ provably controls FDR at level \$q\$.
Moreover, it also appears to have appreciable inferential properties under more
general designs \$X\$ while having substantial power, as demonstrated in a series
of experiments running on both simulated and real data.
@article{bogdan2015slope,
abstract = {We introduce a new estimator for the vector of coefficients \$\beta\$ in the
linear model \$y=X\beta+z\$, where \$X\$ has dimensions \$n\times p\$ with \$p\$
possibly larger than \$n\$. SLOPE, short for Sorted L-One Penalized Estimation,
is the solution to \[\min\_{b\in\mathbb{R}^p}\frac{1}{2}\Vert y-Xb\Vert
\_{\ell\_2}^2+\lambda\_1\vert b\vert \_{(1)}+\lambda\_2\vert
b\vert\_{(2)}+\cdots+\lambda\_p\vert b\vert\_{(p)},\] where
\$\lambda\_1\ge\lambda\_2\ge\cdots\ge\lambda\_p\ge0\$ and \$\vert
b\vert\_{(1)}\ge\vert b\vert\_{(2)}\ge\cdots\ge\vert b\vert\_{(p)}\$ are the
decreasing absolute values of the entries of \$b\$. This is a convex program and
we demonstrate a solution algorithm whose computational complexity is roughly
comparable to that of classical \$\ell\_1\$ procedures such as the Lasso. Here,
the regularizer is a sorted \$\ell\_1\$ norm, which penalizes the regression
coefficients according to their rank: the higher the rank - that is, stronger
the signal - the larger the penalty. This is similar to the Benjamini and
Hochberg [J. Roy. Statist. Soc. Ser. B 57 (1995) 289-300] procedure (BH) which
compares more significant \$p\$-values with more stringent thresholds. One
notable choice of the sequence \$\{\lambda\_i\}\$ is given by the BH critical
values \$\lambda\_{\mathrm {BH}}(i)=z(1-i\cdot q/2p)\$, where \$q\in(0,1)\$ and
\$z(\alpha)\$ is the quantile of a standard normal distribution. SLOPE aims to
provide finite sample guarantees on the selected model; of special interest is
the false discovery rate (FDR), defined as the expected proportion of
irrelevant regressors among all selected predictors. Under orthogonal designs,
SLOPE with \$\lambda\_{\mathrm{BH}}\$ provably controls FDR at level \$q\$.
Moreover, it also appears to have appreciable inferential properties under more
general designs \$X\$ while having substantial power, as demonstrated in a series
of experiments running on both simulated and real data.},
added-at = {2018-12-07T09:10:16.000+0100},
archiveprefix = {arXiv},
author = {Bogdan, Ma{\l}gorzata and van den Berg, Ewout and Sabatti, Chiara and Su, Weijie and Cand\`{e}s, Emmanuel J.},
biburl = {https://www.bibsonomy.org/bibtex/235d7300f44975f4a72574281fdc92401/jpvaldes},
citeulike-article-id = {14027610},
citeulike-linkout-0 = {http://arxiv.org/abs/1407.3824},
citeulike-linkout-1 = {http://arxiv.org/pdf/1407.3824},
citeulike-linkout-2 = {http://dx.doi.org/10.1214/15-aoas842},
citeulike-linkout-3 = {http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4689150/},
citeulike-linkout-4 = {http://view.ncbi.nlm.nih.gov/pubmed/26709357},
citeulike-linkout-5 = {http://www.hubmed.org/display.cgi?uids=26709357},
day = 4,
doi = {10.1214/15-aoas842},
eprint = {1407.3824},
interhash = {9fc31921245a756aa002c9b1316295d7},
intrahash = {35d7300f44975f4a72574281fdc92401},
issn = {1932-6157},
journal = {The Annals of Applied Statistics},
keywords = {owl slope structured_sparsity},
month = nov,
number = 3,
pages = {1103--1140},
pmcid = {PMC4689150},
pmid = {26709357},
posted-at = {2017-11-22 12:42:27},
priority = {2},
timestamp = {2018-12-07T09:40:35.000+0100},
title = {{SLOPE - Adaptive variable selection via convex optimization}},
url = {http://dx.doi.org/10.1214/15-aoas842},
volume = 9,
year = 2015
}