Diagnosing and fixing common problems in Bayesian optimization for molecule design

Austin Tripp, José Miguel Hernández-Lobato
6/11/2024
cs.LGphysics.chem-phstat.ML

No official code found. Best implementation: AustinT/basic-mol-bo-workshop2024 (6 stars, MIT, last commit 2024-06-11).

Abstract

Bayesian optimization (BO) is a principled approach to molecular design tasks. In this paper we explain three pitfalls of BO which can cause poor empirical performance: an incorrect prior width, over-smoothing, and inadequate acquisition function maximization. We show that with these issues addressed, even a basic BO setup is able to achieve the highest overall performance on the PMO benchmark for molecule design (Gao et al 2022). These results suggest that BO may benefit from more attention in the machine learning for molecules community.

View on arXivView PDF

Code Implementations(1)

AustinT/basic-mol-bo-workshop2024Community100%
MIT

Proof of concept molecule BO

61Shell, PythonMay 17, 20242 years agoMIT

Similar Papers

Ranking over Regression for Bayesian Optimization and Molecule Selection

85% match1 repo72024

Best Practices for Multi-Fidelity Bayesian Optimization in Materials and Molecular Research

85% match1 repo172024

A Sober Look at LLMs for Material Discovery: Are They Actually Good for Bayesian Optimization Over Molecules?

84% match1 repo232024

Batched Bayesian optimization by maximizing the probability of including the optimum

83% match1 repo142024

Multi-Variable Batch Bayesian Optimization in Materials Research: Synthetic Data Analysis of Noise Sensitivity and Problem Landscape Effects

83% match2025

Applying Multi-Fidelity Bayesian Optimization in Chemistry: Open Challenges and Major Considerations

83% match1 repo12024

Cite this paper

@article{tripp2024diagnosing,
  title  = {Diagnosing and fixing common problems in Bayesian optimization for molecule design},
  author = {Austin Tripp and José Miguel Hernández-Lobato},
  year   = {2024},
  eprint = {2406.07709},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.07709v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.07709.svg)](https://codeofpaper.com/paper/2406.07709)

Discussion