Diagnosing and fixing common problems in Bayesian optimization for molecule design
Fuente:
arXiv
Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910541011746816 |
|---|---|
| author | Tripp, Austin Hernández-Lobato, José Miguel |
| author_facet | Tripp, Austin Hernández-Lobato, José Miguel |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_07709 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Diagnosing and fixing common problems in Bayesian optimization for molecule design Tripp, Austin Hernández-Lobato, José Miguel Machine Learning Chemical Physics 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. |
| title | Diagnosing and fixing common problems in Bayesian optimization for molecule design |
| topic | Machine Learning Chemical Physics |
| url | https://arxiv.org/abs/2406.07709 |