Ranking over Regression for Bayesian Optimization and Molecule Selection
arXiv:2410.09290 · doi:10.1063/5.0272663
Abstract
Bayesian optimization (BO) has become an indispensable tool for autonomous decision-making across diverse applications from autonomous vehicle control to accelerated drug and materials discovery. With the growing interest in self-driving laboratories, BO of chemical systems is crucial for machine learning (ML) guided experimental planning. Typically, BO employs a regression surrogate model to predict the distribution of unseen parts of the search space. However, for the selection of molecules, picking the top candidates with respect to a distribution, the relative ordering of their properties may be more important than their exact values. In this paper, we introduce Rank-based Bayesian Optimization (RBO), which utilizes a ranking model as the surrogate. We present a comprehensive investigation of RBO's optimization performance compared to conventional BO on various chemical datasets. Our results demonstrate similar or improved optimization performance using ranking models, particularly for datasets with rough structure-property landscapes and activity cliffs. Furthermore, we observe a high correlation between the surrogate ranking ability and BO performance, and this ability is maintained even at early iterations of BO optimization when using ranking surrogate models. We conclude that RBO is an effective alternative to regression-based BO, especially for optimizing novel chemical compounds.
14 + 4 pages, 5 + 3 figures
References in corpus (36)
- Adam: A Method for Stochastic Optimization
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Overcoming catastrophic forgetting in neural networks
- Automatic chemical design using a data-driven continuous representation of molecules
- Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
- On Calibration of Modern Neural Networks
- Fast Graph Representation Learning with PyTorch Geometric
- Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift
- GuacaMol: Benchmarking Models for De Novo Molecular Design
- Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
- GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration
- Open Graph Benchmark: Datasets for Machine Learning on Graphs
- Variational Dropout and the Local Reparameterization Trick
- Accurate Uncertainties for Deep Learning Using Calibrated Regression
- Accelerating high-throughput virtual screening through molecular pool-based active learning
- Bayesian semi-supervised learning for uncertainty-calibrated prediction of molecular properties and active learning
- Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development
- Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective Bayesian Optimization
- Active Ranking using Pairwise Comparisons
- Exploring QSAR Models for Activity-Cliff Prediction
- Roughness of molecular property landscapes and its impact on modellability
- Parallel Bayesian Optimization of Multiple Noisy Objectives with Expected Hypervolume Improvement
- Preferential Bayesian Optimization
- GAUCHE: A Library for Gaussian Processes in Chemistry
- Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks
- Laplace Redux -- Effortless Bayesian Deep Learning
- Gaussian Process Molecule Property Prediction with FlowMO
- Pareto Set Learning for Neural Multi-objective Combinatorial Optimization
- Accelerated Convergence for Counterfactual Learning to Rank
- A benchmark study on reliable molecular supervised learning via Bayesian learning
- Can LLMs Generate Diverse Molecules? Towards Alignment with Structural Diversity
- Learnable Uncertainty under Laplace Approximations
- Aligning Transformers with Continuous Feedback via Energy Rank Alignment
- Pairwise Similarity Learning is SimPLE
- Deep Ranking Ensembles for Hyperparameter Optimization
- Global Human-guided Counterfactual Explanations for Molecular Properties via Reinforcement Learning