bcts's Introduction

Bellman Corrected Tree Search (BCTS)

This repository contains an implementation of the Bellman Corrected Tree Search (BCTS) algorithm, as described in the paper:

Improve Agents without Retraining: Parallel Tree Search with Off-Policy Correction (NeurIPS 2021)

The BCTS algorithm is a search-based method for solving Reinforcement Learning building on Rainbow [Hessel et al., 2017] and NVIDIA CuLE [Dalton et al., 2019].

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

Docker 19 or newer.
Access to NVIDIA Docker Catalog. Visit the NGC website and follow the instructions. This will grant you access to the base docker image (from the Dockerfile) and ability to run on NVIDIA GPU using the nvidia runtime flag.

Installing

Clone the project to get the Dockerfile and build by running

docker build -t bcts .

Usage

Start the docker:

docker run --runtime=nvidia -it bcts /bin/bash

Run the code:
```
cd bcts; python main.py
```
See main.py for optional parameters. For example, for tree depth 2 run:
```
python main.py --tree-depth=2
```

License

This project is licensed under the NVIDIA License.

Acknowledgments

If you use this project please cite:

@article{hallak2021improve,
  title={Improve agents without retraining: Parallel tree search with off-policy correction},
  author={Hallak, Assaf and Dalal, Gal and Dalton, Steven and Froisio, Iuro and Mannor, Shie and Chechik, Gal},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  pages={5518--5530},
  year={2021}
}

Recommend Projects