yaircarmon/recapp

RECAPP: Crafting a More Efficient Catalyst for Convex Optimization

Python implementation of RECAPP and Catalyst for finite sum problems

This repository contains the code to reproduce the experiments from the paper "RECAPP: Crafting a More Efficient Catalyst for Convex Optimization" by Yair Carmon, Arun Jambulapati, Yujia Jin and Aaron Sidford.

Dependencies

To create a conda environment (called recapp) run:

conda env create -f environment.yml  

Running experiments

For examples and explanations on how to run the code, see the notebook:

example.ipynb  

For the (automatically generated) command line interface explanation, run:

python experiment.py algname --help

where algname is either svrg, catalyst, or recapp.

Reference

@inproceedings{carmon2022recapp,
	title={{RECAPP}: Crafting a More Efficient Catalyst for Convex Optimization}}, 
	author={Carmon, Yair and Jambulapati, Arun and Jin, Yujia and Sidford, Aaron},
	booktitle={International Conference on Machine Learning},
	year={2022}
}