Batched Bayesian Optimization by Maximizing the Probability of Including the Optimum.

2024-10-08

Journal of Chemical Information and Modeling

AI Summary

Batch Bayesian optimization (BO) over large-scale chemical libraries faces combinatorial explosion and non-additive batch acquisition functions. Method: We propose qPO (multipoint Probability of Optimality), a batch acquisition strategy that models batch selection as an additive aggregation of single-point optimality probabilities—bypassing combinatorial optimization while implicitly promoting diversity, unlike explicit diversity regularization or parallel Thompson sampling. Leveraging a Gaussian process surrogate, we derive exact analytical gradients of qPO over discrete search spaces and design an efficient greedy batch selection algorithm. Contribution/Results: Experiments demonstrate that qPO matches and complements state-of-the-art batch BO methods in top-1 compound discovery efficiency, significantly enhancing throughput in large-scale molecular screening. qPO establishes a new paradigm for exploitation-oriented discrete BO, offering both theoretical tractability and empirical efficacy.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Batched Bayesian optimization (BO) can accelerate molecular design by efficiently identifying top-performing compounds from a large chemical library. Existing acquisition strategies for batch design in BO aim to balance exploration and exploitation. This often involves optimizing nonadditive batch acquisition functions, necessitating approximation via myopic construction and/or diversity heuristics. In this work, we propose an acquisition strategy for discrete optimization that is motivated by pure exploitation, qPO (multipoint Probability of Optimality). qPO maximizes the probability that the batch includes the true optimum, which is expressed as the sum over individual acquisition scores and thereby circumvents the combinatorial challenge of optimizing a batch acquisition function. We differentiate the proposed strategy from parallel Thompson sampling and discuss how it implicitly captures diversity. Finally, we apply our method to the model-guided exploration of large chemical libraries and provide empirical evidence that it is competitive with and complements other state-of-the-art methods in batched Bayesian optimization.

Problem

• Maximizing probability of including optimum in batch optimization
• Addressing combinatorial challenge in batch acquisition functions
• Enhancing molecular design via efficient compound identification

Innovation

• Maximizes probability of including optimum in batch
• Uses sum of individual scores for efficiency
• Implicitly captures diversity without heuristics