Promoting Stochasticity for Expressive Policies via a Simple and Efficient Regularization Method

This repository is the official implementation of Promoting Stochasticity for Expressive Policies via a Simple and Efficient Regularization Method.

Requirements

To install requirements:

pip install -r requirements.txt

Training

1. You can edit the configuration files in the directory configs.

2. To train agent

python scripts/run.py configs/aceb/aceb_gaussian.json --env_name HalfCheetah-v2

Results

The results are saved in the directory data.