Feature Selection using e-values

This repository contains code and data for the experiments in the paper

Feature Selection using e-values

Subhabrata Majumdar and Snigdhansu Chatterjee

International Conference on Machine Learning (ICML), 2022.

Setup

You should have the R statistical software installed in your system. To install the dependencies, run the following in bash:

Rscript install_dependencies.R

Running the code for the Mixed Integer Optimization (MIO) method requires installation of the Gurobi Optimizer, available under free academic licenses, and the R package by provided by the authors:

Rscript -e "install.packages('devtools')"
Rscript -e "devtools::install_github(repo='ryantibs/best-subset', subdir='bestsubset')"

Contents

The different subdirectories of this repository contain implementation codes for specific experiments. Details are available in README files inside the respective subdirectory.

Subdirectory	Method
lm	Linear models
lmhd	High-dimensional linear models
lmm	Linear mixed models
monsoon	Indian monsoon data analysis
fmri	fMRI data analysis

Contact

Please feel free to email with questions of comments.