Information-Theoretic Safe Exploration with Gaussian Processes

PyTorch implementation of the NeurIPS 2022 paper "Information-Theoretic Safe Exploration with Gaussian Processes". The paper can be found here. The code allows the users to use our implementation of the ISE acquisition function, together with others used in the paper experiment section.

Purpose of the project

This software is a research prototype, solely developed for and published as part of the publication. It will neither be maintained nor monitored in any way.

Setup.

1. Clone the repository and cd into it
2. Create a conda environment ise_exploration with needed dependencies

conda env create --file=environment.yaml

3. Activate the environment
4. Install (locally)

conda activate ise_exploration
pip install -e .

Run example

An example is provided in ise/example.py. To run it, simply activate the environment where ISE has been installed and execute:

python ise/example.py

License

Information-Theoretic Safe Exploration is open-sourced under the AGPL-3.0 license. See the LICENSE file for details.