Memory-Based Dual Gaussian Processes for Sequential Learning
This addresses the problem of accurate sequential learning for practitioners in fields like optimization and continual learning, though it appears incremental as it builds on existing dual sparse variational GP methods.
The paper tackles the challenge of error accumulation in sequential learning with Gaussian processes when past data access is limited, by introducing a method that uses dual sparse variational GPs to control errors and actively update a memory of past data, demonstrating effectiveness in applications like Bayesian optimization, active learning, and continual learning.
Sequential learning with Gaussian processes (GPs) is challenging when access to past data is limited, for example, in continual and active learning. In such cases, errors can accumulate over time due to inaccuracies in the posterior, hyperparameters, and inducing points, making accurate learning challenging. Here, we present a method to keep all such errors in check using the recently proposed dual sparse variational GP. Our method enables accurate inference for generic likelihoods and improves learning by actively building and updating a memory of past data. We demonstrate its effectiveness in several applications involving Bayesian optimization, active learning, and continual learning.