Based on paper "Fighting Boredom in Recommender Systems with Linear Reinforcement Learning", Romain WARLOP, Alessandro Lazaric, Jérémie Mary.
This repository contains:
- an implementation of LinUCRL algorithm;
- preprocessing functions for running on MovieLens-1m dataset.
Requires Python 3.6 and tested on Ubuntu 16.04.
Please check out requirements.txt
for resolving dependency issues.
- Clone this repository
- Download MovieLens dataset
make -B download_dataset
- Run an experiment
make -B train
Check out configuration file lucrl/config/config.yaml
which contains parameters for dataset, mdp and LinUCRL algorithm.