LEAF-GP

This repository contains scripts featured in: .

Please cite this work as:

@article{thebelt2022leafgp,
  title={{Tree ensemble kernels for Bayesian optimization with known constraints over mixed-feature spaces}},
  author={Thebelt, Alexander and Tsay, Calvin and Lee, Robert M and Sudermann-Merx, Nathan and Walz, David and Shafei, Behrang and Misener, Ruth},
  eprint={2207.00879},
  archivePrefix={arXiv},
  year={2022}
}

LEAF-GP Benchmarks

Installation - Linux

Creating a virtual environment

We use virtualenv to setup a virtual environment. You can install this package by running:

python3 -m pip install virtualenv

To set up a new virtual environment called 'env' with Python 3.7 for which this code was tested, run the command:

python3 -m virtualenv env --python=python3.7

in the folder where you want to store the virtual environment. Afterwards, activate the environment using

source env/bin/activate

It is recommended that you update the pip installation in the virtual environment:

pip install --upgrade pip

Install all required packages by running the command:

pip install -r requirements.txt

Installlation of Gurobi 9

Please visit the Gurobi website to receive an academic license and download the solver. To install the optimization modelling environment run:

python -m pip install -i https://pypi.gurobi.com gurobipy

Run black-box function benchmarks

As stated in the paper we evaluate black-box functions: hartmann6d, rastrigin, styblinski_tang, schwefel, g1, g3, g4, g6, g7, g10, alkylation, pressure_vessel and vae_nas. To test LEAF-GP with the hartmann6d benchmark function run:

python run_study.py -bb-func hartmann6d

You can also modify the call by using optional arguments: