Skip to content

Repository for our ICML 2019 paper: Curiosity-Bottleneck

Notifications You must be signed in to change notification settings

whyjay/curiosity-bottleneck

Repository files navigation

Curiosity-Bottleneck:
Exploration by Distilling Task-Specific Novelty

This project hosts the code for our ICML 2019 paper.

Model

Authors

Youngjin Kim, Wontae Nam*, Hyunwoo Kim*, Jihoon Kim, Gunhee Kim (*equal contribution)

Vision and Learning Lab. @ Computer Science and Engineering, Seoul National University
Clova, NAVER

Installation

pip install -r requirements

Run

The following command should train a PPO agent with Curiosity-Bottleneck on Gravitar.

python run_atari.py

Acknowledgements

This code is based on the RND implementation by Yuri Burda.

Citation

@inproceedings{
	kim2019curiositybottleneck,
	title={Curiosity-Bottleneck: Exploration by Distilling Task-Specific Novelty},
	author={Youngjin Kim and Wontae Nam and Hyunwoo Kim and Jihoon Kim and Gunhee Kim},
	booktitle={International Conference on Machine Learning},
	year={2019}
}

About

Repository for our ICML 2019 paper: Curiosity-Bottleneck

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages