A Safe Screening Rule for Sparse Logistic Regression

Authors: Jie Wang, Jiayu Zhou, Jun Liu, Peter Wonka, Jieping Ye

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We have evaluated Slores using high-dimensional data sets from different applications. Experiments demonstrate that Slores outperforms the existing state-of-the-art screening rules and the efficiency of solving sparse logistic regression can be improved by one magnitude.
Researcher Affiliation Collaboration Jie Wang Arizona State University Tempe, AZ 85287 jie.wang.ustc@asu.edu Jiayu Zhou Arizona State University Tempe, AZ 85287 jiayu.zhou@asu.edu Jun Liu SAS Institute Inc. Cary, NC 27513 jun.liu@sas.com Peter Wonka Arizona State University Tempe, AZ 85287 peter.wonka@asu.edu Jieping Ye Arizona State University Tempe, AZ 85287 jieping.ye@asu.edu
Pseudocode Yes Algorithm 1 R = Slores(X, b, λ, λ0, θ λ0)
Open Source Code No The paper does not provide an explicit link or statement about the open-source availability of the code for the Slores method described. It mentions SLEP [14] as a tool used, but not its own implementation.
Open Datasets Yes We evaluate our screening rules using the newgroup data set [10] and Yahoo web pages data sets [23].
Dataset Splits No The paper mentions "undersample 80% of the data" and repeating procedures, but does not explicitly define traditional training/validation/test splits with percentages or sample counts for model development or tuning within a single experimental run.
Hardware Specification Yes All of the experiments are carried out on a Intel(R) (i7-2600) 3.4Ghz processor.
Software Dependencies No Slores, strong rules and SAFE are all implemented in Matlab. All of the experiments are carried out on a Intel(R) (i7-2600) 3.4Ghz processor. The paper mentions MATLAB but does not provide a specific version number. It also mentions SLEP [14] but without a version.
Experiment Setup Yes All of the screening rules are tested along a sequence of 86 parameter values equally spaced on the λ/λmax scale from 0.1 to 0.95. We repeat the procedure 100 times and during each time we undersample 80% of the data.