Parallel Predictive Entropy Search for Batch Global Optimization of Expensive Objective Functions

Abstract

We develop parallel predictive entropy search (PPES), a novel algorithm for Bayesian optimization of expensive black-box objective functions. At each iteration, PPES aims to select a batch of points which will maximize the information gain about the global maximizer of the objective. Well known strategies exist for suggesting a single evaluation point based on previous observations, while far fewer are known for selecting batches of points to evaluate in parallel. The few batch selection schemes that have been studied all resort to greedy methods to compute an optimal batch. To the best of our knowledge, PPES is the first non-greedy batch Bayesian optimization strategy. We demonstrate the benefit of this approach in optimization performance on both synthetic and real world applications, including problems in machine learning, rocket science and robotics.

12 pages in Neural Information Processing Systems 2015

References in corpus (5)

• Practical Bayesian Optimization of Machine Learning Algorithms
• A Tutorial on Bayesian Optimization of Expensive Cost Functions, with Application to Active User Modeling and Hierarchical Reinforcement Learning
• Scalable Bayesian Optimization Using Deep Neural Networks
• Predictive Entropy Search for Efficient Global Optimization of Black-box Functions
• Parallelizing Exploration-Exploitation Tradeoffs with Gaussian Process Bandit Optimization

Cited by in corpus (23)

• Predictive Entropy Search for Multi-objective Bayesian Optimization
• Predictive Entropy Search for Multi-objective Bayesian Optimization with Constraints
• Bayesian Batch Active Learning as Sparse Subset Approximation
• Tuning Hyperparameters without Grad Students: Scalable and Robust Bayesian Optimisation with Dragonfly
• Batched Gaussian Process Bandit Optimization via Determinantal Point Processes
• Multi-fidelity Bayesian Optimization with Max-value Entropy Search and its parallelization
• Is novelty predictable?
• Fast Efficient Hyperparameter Tuning for Policy Gradients
• Parallel Predictive Entropy Search for Multi-objective Bayesian Optimization with Constraints
• GIBBON: General-purpose Information-Based Bayesian OptimisatioN
• Sampling Acquisition Functions for Batch Bayesian Optimization
• Faster Improvement Rate Population Based Training
• BINOCULARS for Efficient, Nonmyopic Sequential Experimental Design
• Near-linear Time Gaussian Process Optimization with Adaptive Batching and Resparsification
• Efficient Batch Black-box Optimization with Deterministic Regret Bounds
• Quantization Algorithms for Random Fourier Features
• Trusted-Maximizers Entropy Search for Efficient Bayesian Optimization
• Bayesian Optimization using Pseudo-Points
• DEEP-BO for Hyperparameter Optimization of Deep Networks
• Robust Policy Search for Robot Navigation
• A tree-based radial basis function method for noisy parallel surrogate optimization
• Parallel Contextual Bandits in Wireless Handover Optimization
• Simple and Scalable Parallelized Bayesian Optimization