Deep Gaussian Process-based Cost-Aware Batch Bayesian Optimization for Complex Materials Design Campaigns

2025-09-17

AI Summary

In complex materials design, the high-dimensional nonlinear search space conflicts with limited experimental resources. Method: We propose a deep Gaussian process (DGP)-driven, cost-aware batch Bayesian optimization framework. It employs a stacked DGP surrogate to explicitly model hierarchical correlations among multiple properties and propagate uncertainty; and introduces a cost-sensitive, heterotopic batch sampling strategy integrated with an extended upper confidence bound (UCB-Ext) acquisition function for efficient parallel recommendation under resource constraints. Results: Evaluated on refractory high-entropy alloy design, the framework converges to optimal compositions in fewer iterations than conventional Gaussian process–based methods. Contribution: This work is the first to unify deep surrogate modeling, uncertainty quantification, and experimental cost constraints within a batch Bayesian optimization paradigm—enhancing both efficiency and robustness in materials inverse design.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The accelerating pace and expanding scope of materials discovery demand optimization frameworks that efficiently navigate vast, nonlinear design spaces while judiciously allocating limited evaluation resources. We present a cost-aware, batch Bayesian optimization scheme powered by deep Gaussian process (DGP) surrogates and a heterotopic querying strategy. Our DGP surrogate, formed by stacking GP layers, models complex hierarchical relationships among high-dimensional compositional features and captures correlations across multiple target properties, propagating uncertainty through successive layers. We integrate evaluation cost into an upper-confidence-bound acquisition extension, which, together with heterotopic querying, proposes small batches of candidates in parallel, balancing exploration of under-characterized regions with exploitation of high-mean, low-variance predictions across correlated properties. Applied to refractory high-entropy alloys for high-temperature applications, our framework converges to optimal formulations in fewer iterations with cost-aware queries than conventional GP-based BO, highlighting the value of deep, uncertainty-aware, cost-sensitive strategies in materials campaigns.

Problem

• Optimizing complex materials design in vast nonlinear spaces
• Efficiently allocating limited evaluation resources cost-effectively
• Modeling hierarchical relationships among high-dimensional compositional features

Innovation

• Deep Gaussian Process surrogate modeling
• Cost-aware batch Bayesian optimization
• Heterotopic querying strategy parallel evaluation