Designing Search Spaces for Unbounded Bayesian Optimization via Transfer Learning (ECML PKDD 2025)

Installation

The easiest way to install the required dependencies is to use Anaconda on Windows (code was tested on Windows 11). In this directory, run

conda env create -f environment.yaml

The environment can then be used with

conda activate transfer_search_space

Alternatively, the dependencies can be installed manually using environment.yaml as reference.

Experiments

The experiments contain in the file test_proposed.py

Example: Running Scenario 1 with Ackley function

python test_proposed.py --test_func ackley_s1 --restart 20 --max_evals 50 --save_result False