Batched Gaussian Process Bandit Optimization via Determinantal Point Processes

Abstract

Gaussian Process bandit optimization has emerged as a powerful tool for optimizing noisy black box functions. One example in machine learning is hyper-parameter optimization where each evaluation of the target function requires training a model which may involve days or even weeks of computation. Most methods for this so-called "Bayesian optimization" only allow sequential exploration of the parameter space. However, it is often desirable to propose batches or sets of parameter values to explore simultaneously, especially when there are large parallel processing facilities at our disposal. Batch methods require modeling the interaction between the different evaluations in the batch, which can be expensive in complex scenarios. In this paper, we propose a new approach for parallelizing Bayesian optimization by modeling the diversity of a batch via Determinantal point processes (DPPs) whose kernels are learned automatically. This allows us to generalize a previous result as well as prove better regret bounds based on DPP sampling. Our experiments on a variety of synthetic and real-world robotics and hyper-parameter optimization tasks indicate that our DPP-based methods, especially those based on DPP sampling, outperform state-of-the-art methods.

To appear at NIPS 2016

References in corpus (7)

• Practical Bayesian Optimization of Machine Learning Algorithms
• Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design
• Predictive Entropy Search for Efficient Global Optimization of Black-box Functions
• Batch Bayesian Optimization via Local Penalization
• Parallel Predictive Entropy Search for Batch Global Optimization of Expensive Objective Functions
• GLASSES: Relieving The Myopia Of Bayesian Optimisation
• Fast DPP Sampling for Nyström with Application to Kernel Methods

Cited by in corpus (19)

• Optimal Transport Kernels for Sequential and Parallel Neural Architecture Search
• Pareto Front-Diverse Batch Multi-Objective Bayesian Optimization
• Parallel Predictive Entropy Search for Multi-objective Bayesian Optimization with Constraints
• PHOENICS: A universal deep Bayesian optimizer
• GIBBON: General-purpose Information-Based Bayesian OptimisatioN
• Sampling Acquisition Functions for Batch Bayesian Optimization
• Parallelizing Thompson Sampling
• Open Loop Hyperparameter Optimization and Determinantal Point Processes
• Near-linear Time Gaussian Process Optimization with Adaptive Batching and Resparsification
• On Batch Bayesian Optimization
• A Hierarchical Approach to Scaling Batch Active Search Over Structured Data
• DPPy: Sampling DPPs with Python
• Efficient Phase Diagram Sampling by Active Learning
• DEEP-BO for Hyperparameter Optimization of Deep Networks
• Designing over uncertain outcomes with stochastic sampling Bayesian optimization
• Simple and Scalable Parallelized Bayesian Optimization
• Sampling from a $k$-DPP without looking at all items
• Bayesian optimization for modular black-box systems with switching costs
• Scaling Bayesian Optimization With Game Theory