A Unifying View of Sparse Approximate Gaussian Process Regression

By Joaquin Quiñonero Candela, Carl E. Rasmussen. 2005

2,213 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Probabilistic & Bayesian Methods

Citations in the corpus

Zero-shot Knowledge Transfer via Adversarial Belief Matching (2019, methodology): “While early techniques used a subset of the training data as inducing points (Candela and Rasmussen, 2005), creating pseudo data using variational techniques was later shown more efficient (Titsias, 2009).”
Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, methodology): “All sparse pseudoinput GP approximations can be specified in terms of the form used for the training and test conditionals qðf j uÞ and qðf ? j uÞ, respectively [124].”
Distinguishing Cause from Effect Using Observational Data: Methods and Benchmarks (2014, methodology): “We used the FITC approximation (Quiñonero-Candela and Rasmussen, 2005) as an approximation for exact GP regression in order to reduce computation time.”
Variational Auto-Regressive Gaussian Processes for Continual Learning (2020, methodology): “This section focuses on pseudo-point based methods [23, 4] and, in particular, the seminal work of Titsias [36], Hensman et al.”
Functional Regularisation for Continual Learning using Gaussian Processes (2019, methodology): “by memorising an approximate posterior belief over the underlying task-speciﬁc function. To implement this, we consider Gaussian processes (GPs) [23], and make use of inducing point sparse GP methods [6, 16, 33, 22, 35, 36, 12, 4], which summarise posterior distributions over functions using small n”
Deep Kernel Learning (2015, methodology): “KISS-GP training scales as O(n+ h(m)) (where h(m) is typically close to linear in m), versus conventional scalable GP approaches which require O(m2n + m(3)) (Quiñonero-Candela and Rasmussen, 2005) computations and need m n for tractability, which results in severe deteriorations in predictive perfor”
Stochastic Variational Deep Kernel Learning (2016, background): “Many promising approaches to scalability have been explored, for example, involving randomized methods [20, 16, 31] , and low rank approximations [23, 19].”
Scalable Variational Gaussian Process Classification (2014, methodology): “However we might expect better performance by relaxing the condition that the inducing points be a sub-set of the data, as is the case for regression [Qui˜nonero-Candela and Rasmussen, 2005].”
Streaming Sparse Gaussian Process Approximations (2017, background): “This prohibitive complexity of exact learning and inference in GP models has driven the development of many sparse approximation frameworks [12, 13].”