Product Kernel Interpolation for Scalable Gaussian Processes

Abstract

Recent work shows that inference for Gaussian processes can be performed efficiently using iterative methods that rely only on matrix-vector multiplications (MVMs). Structured Kernel Interpolation (SKI) exploits these techniques by deriving approximate kernels with very fast MVMs. Unfortunately, such strategies suffer badly from the curse of dimensionality. We develop a new technique for MVM based learning that exploits product kernel structure. We demonstrate that this technique is broadly applicable, resulting in linear rather than exponential runtime with dimension for SKI, as well as state-of-the-art asymptotic complexity for multi-task GPs.

arXiv commentAppears in Artificial Intelligence and Statistics (AISTATS) 21, 2018

References in corpus (10)

• Practical Bayesian Optimization of Machine Learning Algorithms
• Structure Discovery in Nonparametric Regression through Compositional Kernel Search
• A Framework for Individualizing Predictions of Disease Trajectories by Exploiting Multi-Resolution Structure
• Product Kernel Interpolation for Scalable Gaussian Processes
• A Bayesian Nonparametric Approach for Estimating Individualized Treatment-Response Curves
• Scalable Log Determinants for Gaussian Process Kernel Learning
• Sparse Multi-Output Gaussian Processes for Medical Time Series Prediction
• Improving stochastic estimates with inference methods: calculating matrix diagonals
• Additive Kernels for Gaussian Process Modeling
• Improved stochastic trace estimation using mutually unbiased bases

Cited by in corpus (15)

• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration
• When Gaussian Process Meets Big Data: A Review of Scalable GPs
• Product Kernel Interpolation for Scalable Gaussian Processes
• Scaling Gaussian Process Regression with Derivatives
• Exact Gaussian Processes on a Million Data Points
• Kernel methods through the roof: handling billions of points efficiently
• Constant-Time Predictive Distributions for Gaussian Processes
• Fast Matrix Square Roots with Applications to Gaussian Processes and Bayesian Optimization
• Sparse Orthogonal Variational Inference for Gaussian Processes
• Scalable Gaussian Process Classification with Additive Noise for Various Likelihoods
• Fast Kronecker Matrix-Matrix Multiplication on GPUs
• Scalable Gaussian Process Inference with Finite-data Mean and Variance Guarantees
• The Renyi Gaussian Process: Towards Improved Generalization
• Faster Kernel Interpolation for Gaussian Processes