Lecture 12

Bayesian Optimization

Surrogate models and acquisition functions — closing the loop between prediction and experiment for efficient molecular discovery.

Jules Schleinitz
90 min
Back to schedule
Slides (PDF) Slides (PPTX) Recording (YouTube) Companion Tutorial



Shields et al. — "Bayesian Reaction Optimization as a Tool for Chemical Synthesis" Paper
Nature, 2021, 590, 89–96 · EDBO; the Suzuki-coupling human-vs-BO benchmark used throughout this lecture
Shahriari et al. — "Taking the Human Out of the Loop: A Review of Bayesian Optimization" Paper
Proc. IEEE, 2016, 104(1), 148–175 · The standard BO review article
Srinivas et al. — "Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting" Paper
IEEE Trans. Inf. Theory, 2012, 58(5), 3250–3265 · The GP-UCB acquisition function
MacLeod et al. — "Self-Driving Laboratory for Accelerated Discovery of Thin-Film Materials" Paper
Sci. Adv., 2020, 6(20), eaaz8867 · Materials example: autonomous BO of hole mobility for perovskite solar cells
Graff, Shakhnovich & Coley — "Accelerating High-Throughput Virtual Screening through Molecular Pool-Based Active Learning" Paper
Chem. Sci., 2021, 12, 7866–7881 · Drug-discovery example: BO-guided screening over 100M+ compound libraries
Garnett — "Bayesian Optimization" Book
Cambridge University Press, 2023 · bayesoptbook.com · The definitive textbook — full PDF free for personal use
BoTorch Documentation Website
botorch.org · PyTorch-based BO library — implements q-EI, q-NEHVI, and the batch/multi-objective methods covered in this lecture
Honegumi — BO script generator Tool
honegumi.readthedocs.io · Pick your problem shape (multi-objective, categorical, batch, constrained) and it emits working Ax/BoTorch code — a fast way to get past the boilerplate after this lecture