BEACON: A Bayesian Optimization Inspired Strategy for Efficient Novelty Search

This repo contains the codes for BEACON: A Bayesian Optimization Inspired Strategy for Efficient Novelty Search

Installation

pip install -r requirements.txt

Usage

We provide the code scripts for executing BEACON on different problem setting. Noted that all code script requires the usage of ThompsonSampling.py file to perform efficient Thompson sampling strategy proposed in this work.

Running Experiments

Run the following commands to execute BEACON under different problem setting:

Continuous feature space (e.g. synthetic problem conducted in this paper)

For continuous feature space and single outcome problem:

python Continuous_SingleOutcome_BEACON.py

For continuous feature space and multi outcome problem:

python Continuous_Multioutcome_BEACON.py

Discrete feature space (e.g. Material and drug case study conducted in this paper)

For discrete feature space and single outcome problem:

python Discrete_SingleOtcome_BEACON.py

For discrete feature space and multi outcome problem:

python Discrete_MultiOutcome_BEACON.py

Extension to high-dimensional problem

We propose an extension algorithm TR-BEACON for addressing high-dimensional novelty search problem. Please refer to the paper and github repo.