The Non-Overlapping Statistical Approximation to Overlapping Group Lasso
Authors: Mingyu Qi, Tianxi Li
JMLR 2024 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The effectiveness of our method is demonstrated through extensive simulation examples and a predictive task of cancer tumors. ... Empirical evaluations using both simulated and real breast cancer gene expression data are presented in Sections 5 and 6, respectively. |
| Researcher Affiliation | Academia | Mingyu Qi EMAIL Department of Statistics University of Virginia Charlottesville, VA 22904, USA Tianxi Li EMAIL School of Statistics University of Minnesota, Twin Cities Minneapolis, MN 55455, USA |
| Pseudocode | Yes | Algorithm 1 BCD algorithm for the proximal operator of the overlapping group lasso ... Algorithm 2 Algorithm to construct the overlapping-induced partition G |
| Open Source Code | No | Two MATLAB-based solvers are employed for the overlapping group lasso problems. The first solver, Fo GLasso (Yuan et al., 2011), is from the SLEP package (Liu et al., 2009b). It can handle general overlapping group structures. The second solver, from the SPAM package (Mairal et al., 2014)... SLEP and SPAM package solvers were also applied to solve lasso and non-overlapping group lasso estimators in our benchmark set to ensure that the timing comparison implementation is consistent. |
| Open Datasets | Yes | We use the gene expression data from Van De Vijver et al. (2002) as the covariate matrix X, which can be accessed through the R package breast Cancer NKI (Schroeder et al., 2021). ... five gene pathway sets from the Molecular Signatures Database (Subramanian et al., 2005) as group structures, summarized in Table 1. |
| Dataset Splits | Yes | We adopt the evaluation procedure of Lee and Xing (2014), where we randomly split the data set into 200 training observations and 95 test observations. All methods are tuned by 5-fold crossvalidation on the training data. |
| Hardware Specification | No | The authors acknowledge the Minnesota Supercomputing Institute (MSI) at the University of Minnesota and the Research Computing at the University of Virginia for providing resources that contributed to the research results reported within this paper. |
| Software Dependencies | Yes | Two MATLAB-based solvers are employed for the overlapping group lasso problems. The first solver, Fo GLasso (Yuan et al., 2011), is from the SLEP package (Liu et al., 2009b). ... The second solver, from the SPAM package (Mairal et al., 2014)... R package breast Cancer NKI (Schroeder et al., 2021). R package version 1.32.0. |
| Experiment Setup | Yes | The group weight in the overlapping group lasso problem is wg = p/dg, as is usually used in practice. For all methods, we employ the absolute difference in function values between iterations as the stopping criterion, with a tolerance set at 10-5. ... All methods are tuned by 5-fold crossvalidation on the training data. |