kiminh / dpprl

dpprl

A DPP implementation on Deep Reinforcement Learning for Movies Recommendation System from the paper:

Zhao, X., Zhang, L., Xia, L., Ding, Z., Yin, D., & Tang, J. (2019). Deep reinforcement learning for list-wise recommendations.

The implementation is heavily built on the following code repo:

egipcy. (2020). List-wise recommendation framework based on deep reinforcement learning. https://github.com/egipcy/LIRD. GitHub.

This decision was made to run the experiment in the available time, and to built on top of a stable fundament.

Requirements

• Python 3
• Tensorflow 1.15
• Numpy
• Scipy
• Dppy

Usage

Install required packages
run python deep-dpprl-dpp.py

Configurations

The dataset is split into train, validation, and test sets to make the experiment as valid as possible.
By tuning the nb_episodes parameter the model is tested on diversifying performance through the episodes.

Acknowledgments

This project is heavily built on the LIRD model, and by using the following papers:

• egipcy. (2020). List-wise recommendation framework based on deep reinforcement learning. https://github.com/egipcy/LIRD. GitHub.
• Zhao, X., Zhang, L., Xia, L., Ding, Z., Yin, D., & Tang, J. (2019). Deep reinforcement learning for list-wise recommendations.
• Gautier, G., Polito, G., Bardenet, R., & Valko, M. (2019). Dppy: Dpp sampling with python. Journal of Machine Learning Research, 20(180), 1-7.
• Liu, Y., Zhang, Y., Wu, Q., Miao, C., Cui, L., Zhao, B., . . . Guan, L. (2019). Diversity-promoting deep reinforcement learning for interactive recommendation. ArXiv, abs/1903.07826 .