/wu-go

Official code for the paper Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space.

Primary LanguagePythonMIT LicenseMIT

Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space

Official code for the paper Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space.

Tigran Ramazyan*, Mikhail Hushchyn, Denis Derkach.

Pypi package

WAGGON: WAsserstein Global Gradient-free OptimisatioN

Installation

  • Create conda environment:
conda create -n wugo python=3.10
conda activate wugo
  • Install core dependencies:
pip install -r requirements.txt

Experiments

  • To run experiments:
    python opt_exp.py experiment acquisition_function surrogate_model

Available options:

  • experiment: three_hump_camel, ackley, levi, himmelblau, rosenbrock8, rosenbrock20, tang
  • acquisition_function: WU-GO, EI, LCB
  • surrogate_model: GAN, BNN, DE, GP, DGP

Citation

@misc{ramazyan2024globaloptimisationblackboxfunctions,
      title={Global Optimisation of Black-Box Functions with Generative Models in the Wasserstein Space}, 
      author={Tigran Ramazyan and Mikhail Hushchyn and Denis Derkach},
      year={2024},
      eprint={2407.11917},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2407.11917}, 
}