GAP Safe Screening Rules for Sparse-Group Lasso

Authors: Eugene Ndiaye, Olivier Fercoq, Alexandre Gramfort, Joseph Salmon

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section we present our experiments and illustrate the numerical benefit of screening rules for the Sparse-Group Lasso.
Researcher Affiliation Academia Eugene Ndiaye, Olivier Fercoq, Alexandre Gramfort, Joseph Salmon LTCI, CNRS, Télécom Paris Tech Université Paris-Saclay 75013 Paris, France first.last@telecom-paristech.fr
Pseudocode Yes Algorithm 1 Computation of Λpx, α, Rq.
Open Source Code Yes The source code can be found in https://github.com/Eugene Ndiaye/GAPSAFE_SGL.
Open Datasets Yes Real dataset: NCEP/NCAR Reanalysis 1 [14]
Dataset Splits Yes We choose τ in the set t0, 0.1, . . . , 0.9, 1u by splitting in 50% the observations and run a training-test validation procedure.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts, or detailed computer specifications) were provided for the experimental setup.
Software Dependencies No No specific software dependencies with version numbers (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) were provided. Mentions include 'R glmnet package' and the 'ISTA-BC algorithm' without versions.
Experiment Setup Yes By default, we choose δ 3 and T 100, following the standard practice when running crossvalidation using sparse models (see R glmnet package [11]). The weights are always chosen as wg ?ng (as in [17]). The expensive computation of the dual gap is not performed at each pass over the data, but only every f ce pass (in practice f ce 10 in all our experiments).