Ridge Regression and Provable Deterministic Ridge Leverage Score Sampling

This repository has been developed to reproduce the results in McCurdy (2018, http://arxiv.org/abs/1803.06010). We implement the deterministic ridge leverage sampling algorithm that comes with a $(1 + \epsilon)$ error column subset selection, $(1 + \epsilon)$ error projection-cost preservation, an additive-multiplicative spectral bound, and a $(1 + \epsilon)$ bound on the statistical risk for ridge regression. The code in this repository will download the biological data for you, perform the analysis, and make the figures.

Installing the dependencies

To download the TCGA lower-grade glioma (LGG) tumor multi-omic data collected by the TCGA Research Network, you will first need to install CNTools (Zhang, 2015) and TCGA2STAT (Wan et al., 2016) in R. To install the R packages, open an R environment and type the following:

## try http:// if https:// URLs are not supported
source("https://bioconductor.org/biocLite.R")
biocLite("CNTools")
install.packages("TCGA2STAT_1.0.tar.gz", repos = NULL, type = "source")

The rest of this code repository is in python. We used the anaconda package manager which can be obtained at https://www.continuum.io/downloads, and we include a .yml file (ridge.yml) that we use for the analysis. You can simply create an "conda env" with the .yml file by

conda env create -f ridge.yml 
#
# To activate this environment, use:
# $ source activate ridge
#
# To deactivate this environment, use:
# $ source deactivate
#

Getting Started

In order to run the code, first clone this repository.

git clone https://github.com/srmcc/determinstic-ridge-leverage-sampling.git

Then activate your conda environment:

source activate ridge

Change directory to the cloned repository:

cd  /path/to/determinstic-ridge-leverage-sampling

And run the analysis file.

python ridge.py