Parallel and Distributed Thompson Sampling for Large-scale Accelerated Exploration of Chemical Space

2017/06/06 by José Miguel Hernández-Lobato, Hernández-Lobato, José Miguel, James Requeima +7 · 20 citations
Computer Science · Decision Sciences · Engineering · Mathematics · #Advanced Bandit Algorithms Research #Advanced Multi-Objective Optimization Algorithms #FOS: Computer and information sciences #Fault Detection and Control Systems #Machine Learning (stat.ML) #Machine Learning and Algorithms #Machine Learning and Data Classification #stat.ML

paper · pdf · doi:10.48550/arxiv.1706.01825

Accepted for publication in the proceedings of the 2017 ICML conference

arxiv created 2017/06/06 · openalex publication_date 2017/06/06 · arxiv updated 2017/06/07 · openalex created_date 2022/09/30 · openalex updated_date 2026/07/28

Abstract

Chemical space is so large that brute force searches for new interesting molecules are infeasible. High-throughput virtual screening via computer cluster simulations can speed up the discovery process by collecting very large amounts of data in parallel, e.g., up to hundreds or thousands of parallel measurements. Bayesian optimization (BO) can produce additional acceleration by sequentially identifying the most useful simulations or experiments to be performed next. However, current BO methods cannot scale to the large numbers of parallel measurements and the massive libraries of molecules currently used in high-throughput screening. Here, we propose a scalable solution based on a parallel and distributed implementation of Thompson sampling (PDTS). We show that, in small scale problems, PDTS performs similarly as parallel expected improvement (EI), a batch version of the most widely used BO heuristic. Additionally, in settings where parallel EI does not scale, PDTS outperforms other scalable baselines such as a greedy search, ε-greedy approaches and a random search method. These results show that PDTS is a successful solution for large-scale parallel BO.

Citations

• Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks
• Density-functional thermochemistry. III. The role of exact exchange
• Batch Bayesian Optimization via Local Penalization

Cited by

• Warm-Starting Iterative Gaussian Processes for Faster Sequential Inference
• Thompson Sampling via Fine-Tuning of LLMs
• RAISE: A self-driving laboratory for interfacial property formulation discovery
• Scalable Constrained Bayesian Optimization
• Practical Bayesian Optimization of Objectives with Conditioning Variables
• Bayesian Graph Neural Networks for Molecular Property Prediction
• Gradient-based grand canonical optimization enabled by graph neural networks with fractional atomic existence
• GPU Accelerated Exhaustive Search for Optimal Ensemble of Black-Box Optimization Algorithms
• Sampling from Gaussian Processes: A Tutorial and Applications in Global Sensitivity Analysis and Optimization
• Actively Learning what makes a Discrete Sequence Valid
• Bayesian Screening: Multi-test Bayesian Optimization Applied to in silico Material Screening
• New Heuristics for Parallel and Scalable Bayesian Optimization
• Myopic Bayesian Design of Experiments via Posterior Sampling and Probabilistic Programming
• Feasibility-Driven Trust Region Bayesian Optimization
• Simplifying Bayesian Optimization Via In-Context Direct Optimum Sampling
• Bencher: Simple and Reproducible Benchmarking for Black-Box Optimization
• Improving LLM-based Global Optimization with Search Space Partitioning
• Robust and Computation-Aware Gaussian Processes
• Bayesian Variational Optimization for Combinatorial Spaces
• Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project
• Towards "on-demand" van der Waals epitaxy with hpc-driven online ensemble sampling

Related

• HypBO: Accelerating Black-Box Scientific Experiments Using Experts' Hypotheses
• Informing Acquisition Functions via Foundation Models for Molecular Discovery
• Asynchronous Parallel Bayesian Optimisation via Thompson Sampling
• Practical Massively Parallel Monte-Carlo Tree Search Applied to Molecular Design
• Thompson Sampling via Fine-Tuning of LLMs