TUHH-ICS / 2024-code-L4DC-Towards-Safe-Multi-Task-Bayesian-Optimization

Code for paper "Safe Multi-Task Bayesian Optimization"

Towards Safe Multi-Task Bayesian Optimization

General

This repository contains supplementary material and the code to reproduce the tables and figures presented in

J. O. Lübsen, C. Hespe, A. Eichler, "Towards Safe Multi-Task Bayesian Optimization", submitted to the 6th Annual Learning for Dynamics and Control Conference, 2024

The preprint with supplementary material including proofs is available on ArXiv:

http://arxiv.org/abs/2312.07281

The code has three main entry points, which are located in the code directory. Usually, the user needs to do some adjustments which are specified in the respective file.

1. test_run.py starts the optimization of a low dimensional problem with additional online generated figures. It provides a good illustration of how the algorithm works. In the file, the user can switch between a one and two dimensional problem.
2. run_N2.py starts to generate the data that is used in Figure 3 (a). The script needs to be executed repetitively with different disturbances.
3. run_N5.py starts to generate the data similar to Figure 3 (b) (only SaMSBO).

Note that running scripts run_N2.py and run_N5.py may take a long time. After generating the data, the script plot.py can be used to plot a figure similar to Figure 3.

Prerequisites

To run the code install python3.10 and the dependencies specified in requirements.txt.

pip install -r requirements.txt

The code in this repository was tested in the following environment:

• *Ubuntu 22.04.3 LTS
• *Python 3.10.12