Safe Screening with Variational Inequalities and Its Application to Lasso

Authors: Jun Liu, Zheng Zhao, Jie Wang, Jieping Ye

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on both synthetic and real data sets are reported to demonstrate the effectiveness of the proposed Sasvi for Lasso screening.
Researcher Affiliation Collaboration Jun Liu, Zheng Zhao {JUN.LIU,ZHENG.ZHAO}@SAS.COM SAS Institute Inc., Cary, NC 27513 Jie Wang, Jieping Ye {JIE.WANG.USTC,JIEPING.YE}@ASU.EDU Arizona State University, Tempe, AZ 85287
Pseudocode No The paper contains mathematical derivations and theoretical explanations, but no structured pseudocode or algorithm blocks are provided.
Open Source Code No The paper states: "For the Lasso solver, we make use of the SLEP package (Liu et al., 2009). URL http://www.public.asu.edu/~jye02/Software/SLEP." This link is to a third-party solver used by the authors, not the open-source code for the Sasvi method itself.
Open Datasets Yes PIE Face Image Data Set The PIE face image data set used in this experiment 4 contains 11554 gray face images... 4http://www.cad.zju.edu.cn/home/dengcai/ Data/Face Data.html. MNIST Handwritten Digit Data Set This data set contains grey images of scanned handwritten digits, including 60, 000 for training and 10, 000 for testing.
Dataset Splits No The paper describes how synthetic data is generated and how the PIE and MNIST datasets are used to construct X and y. For example, for MNIST: "we first randomly select 5000 images for each digit from the training set (and in total we have 50000 images) and get a data matrix X R784 50000, and then we randomly select an image from the testing set and treat it as the response vector y R784." While it uses parts of standard datasets, it does not specify explicit training/validation/test splits for the models being evaluated (e.g., percentages or sample counts for each split).
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments, such as GPU models, CPU specifications, or memory.
Software Dependencies No The paper mentions using the "SLEP package (Liu et al., 2009)" as the Lasso solver. However, it does not specify a version number for SLEP or any other software dependencies, such as programming languages or libraries with their versions.
Experiment Setup Yes For a given generated data set (X and y), we run the solver with or without screening rules to solve the Lasso problems along a sequence of 100 parameter values equally spaced on the λ/λmax scale from 0.05 to 1.0. The reported results are averaged over 100 trials of randomly drawn X and y.