Frank-Wolfe methods for Recommender Systems

The final project created for Optimization for Data Science course.

The project covers application of two important algorithms designed for constraint optimization tasks: Frank-Wolfe and Projected Gradient.

Introduced code is aimed to demonstrate the algorithms' performance for real world Recommender System task. In order to do that, we have selected Netflix Prize as such a task, formalized it as constrained and convex optimization problem, applied the algorithms to that problem.

Contributors

Kokot Maksim

UNIPD, 2023

Usage

1. Download the real dataset here https://files.grouplens.org/datasets/movielens/ml-latest-small.zip
2. Put ratings.csv into data folder
3. Use the code provided in notebook folder

Project Organization

├── README.md                        <- The top-level README for developers using this project.
├── data                             <- Folder for real dataset.
├── notebooks                        <- Jupyter notebooks folder containing tho code.
├── report                           <- Folder containing report pdf file
└── requirements.txt                 <- The requirements file for reproducing the analysis environment, e.g.
                                        generated with `pip freeze > requirements.txt`