Constrained Optimization using Frank-Wolfe Type Algorithm

This repository provides MATLAB code for performing constrained optimization using Frank-Wolfe type algorithms. The code is organized into a package, and several configurable scripts are provided to facilitate different types of analyses and comparisons.

Available Scripts

test.m

This script allows you to perform and evaluate a Frank-Wolfe type algorithm, either "Standard" or "Away-Step". You can configure the script to suit your specific needs.

benchmark.m

This script enables comparison between all selected variants of the Frank-Wolfe algorithm and selected off-the-shelf algorithms from quadprog ("interior-point-convex" or "active-set").

expe_test.m

This script performs multiple runs of a selected Frank-Wolfe type algorithm, varying the parameter values based on provided lists.

expe_benchmark.m

This script conducts multiple comparisons between all selected Frank-Wolfe type algorithms and off-the-shelf methods, varying the parameter values based on provided lists.

Code Structure

All function codes are located in the src folder. To use these functions in the executable scripts, add the source folder to the MATLAB path using the command addpath src.

Results from each experiment are automatically saved in a sub-folder within the results folder. The sub-folder is named based on the date of the experiment.

Getting Started

To run the experiments and test the code:

1. Launch MATLAB and navigate to the main folder of this repository.
2. Open one of the configurable scripts (test.m, benchmark.m, expe_test.m, or expe_benchmark.m).
3. Configure the parameters within the script as needed.
4. Run the script.

Explore and modify the scripts as needed to suit your specific use case. Enjoy optimizing!