mlsanityData to 13 Oct 2026

Diagnosing and fixing common problems in Bayesian optimization for molecule design

arXiv 2406.07709 ↗First posted 11 Jun 2024cs.LG physics.chem-ph stat.MLcited by 9

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. Cited by 9 papers in the map (data through 13 Oct 2026). Builds on 6 of its 12 references that are in the map. First posted on arXiv 11 Jun 2024.

Builds on

6 of 12 references.

References · 6

  1. 2021Flow Network based Generative Models for Non-Iterative Diverse Candidate Generationcited by 386
  2. 2021GFlowNet Foundationscited by 286
  3. 2017Molecular De Novo Design through Deep Reinforcement Learningcited by 264
  4. 2022Sample Efficiency Matters: A Benchmark for Practical Molecular Optimizationcited by 151
  5. 2019Augmenting Genetic Algorithms with Deep Neural Networks for Exploring the Chemical Spacecited by 66
  6. 2023Genetic algorithms are strong baselines for molecule generationcited by 31

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2024: 1; 2025: 5; 2026: 3.

9 papers.

  1. 2025Adaptive Kernel Design for Bayesian Optimization Is a Piece of CAKE with LLMscited by 16
  2. 2025Return of the Latent Space COWBOYS: Re-thinking the use of VAEs for Bayesian Optimisation of Structured Spacescited by 13
  3. 2024Small Molecule Optimization with Large Language Modelscited by 5
  4. 2025A Genetic Algorithm for Navigating Synthesizable Molecular Spacescited by 4
  5. 2026Sample Efficient Generative Optimization for Molecular Designcited by 1
  6. 2025Why risk matters for protein binder designcited by 0
  7. 2025A study of EHVI vs fixed scalarization for molecule designcited by 0
  8. 2026MolLIBRA: Genetic Molecular Optimization with Multi-Fingerprint Surrogates and Text-Molecule Aligned Criticcited by 0
  9. 2026Scientific Discovery under Validation Congestion via Multi-Fidelity Pairwise Rankingscited by 0