Rapid optimization in high dimensional space by deep kernel learning augmented genetic algorithms

2024-10-04

arXiv.org

AI Summary

Genetic algorithms (GAs) suffer from prohibitively high evaluation costs in high-dimensional complex-space optimization (e.g., molecular discovery, battery charging policy design), while deep kernel learning (DKL)–based models lack inherent generative capability for structural candidate synthesis. Method: We propose the first hybrid GA-DKL optimization framework, synergistically integrating GA’s structural generation capacity with a DKL-enhanced Gaussian process surrogate model to enable rapid prediction and iterative refinement of novel candidates; the framework is natively compatible with Bayesian optimization pipelines. Contribution/Results: On the FerroSIM model optimization task, our method substantially reduces computational overhead. Across diverse domains—including molecular design and energy systems—it demonstrates superior optimization efficiency and strong cross-domain generalization. By unifying generative modeling with predictive accuracy, this work establishes a novel paradigm for high-dimensional black-box optimization that simultaneously addresses synthesis and surrogate-based inference.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders
• Constraint Satisfaction and Optimization: Mixed Discrete/Continuous Optimization

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Exploration of complex high-dimensional spaces presents significant challenges in fields such as molecular discovery, process optimization, and supply chain management. Genetic Algorithms (GAs), while offering significant power for creating new candidate spaces, often entail high computational demands due to the need for evaluation of each new proposed solution. On the other hand, Deep Kernel Learning (DKL) efficiently navigates the spaces of preselected candidate structures but lacks generative capabilities. This study introduces an approach that amalgamates the generative power of GAs to create new candidates with the efficiency of DKL-based surrogate models to rapidly ascertain the behavior of new candidate spaces. This DKL-GA framework can be further used to build Bayesian Optimization (BO) workflows. We demonstrate the effectiveness of this approach through the optimization of the FerroSIM model, showcasing its broad applicability to diverse challenges, including molecular discovery and battery charging optimization.

Problem

• Combines genetic algorithms with deep kernel learning for optimization
• Reduces computational cost in high-dimensional space exploration
• Enables efficient candidate evaluation in complex optimization problems

Innovation

• Combines genetic algorithms with deep kernel learning
• Uses surrogate models to evaluate candidate solutions efficiently
• Enables Bayesian optimization workflows for diverse applications