/FSAF

Primary LanguagePythonGNU Affero General Public License v3.0AGPL-3.0

FewShotAF

This repository contains implementations of the paper, Reinforced Few-Shot Acquisition Function Learning for Bayesian Optimization. It includes code for running the training and evaluation all experiments in the paper. To get our initial model, run python train_fsaf.py. To comparison with other afs, we implemented evaluate_fsaf_blackbox.py for global optimization benchmark functions, evaluate_fsaf_gps.py for GP kernel functions and evaluate_fsaf_hpo.py for real task and hpo.

Quick run

Examples of FSAF running script can see in the run_all_exp.sh We tested this code on the following versions of libraries:

  • gpy 1.9.9
  • pytorch 1.7.0
  • numpy 1.18.1
  • python 3.7.6
  • scipy 1.4.1
  • sympy 1.5.1
  • sobol-seq 0.1.2
  • gym 0.17.1
  • namedlist 1.7
  • matplotlib 3.1.3

Refernece