Sparse Spectrum Gaussian Process Regression

By Miguel L'azaro-Gredilla, Joaquin Quiñonero Candela, Carl E. Rasmussen, Aníbal R. Figueiras-Vidal. 2010

546 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Control and Systems Engineering

How researchers see this paper

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

In Artificial Intelligence

In Computational Theory and Mathematics

In Control and Systems Engineering

Citations in the corpus

Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, background): “2) Sparse spectrum Gaussian processes (SSGP): While inducing pseudo-inputs reduce computational complexity by using a ﬁxed number of points in the search space, sparse spectrum Gaussian processes (SSGP) take a similar approach to the kernel’s spectral space [94].”
Calibrating Deep Convolutional Gaussian Processes (2018, methodology): “In particular, we propose to replace the fully connected layers of CNN s with GP s that we approximate with random features [4, 19].”
Agile Autonomous Driving using End-to-End Deep Imitation Learning (2017, methodology): “In this work we consider a statistical model, Sparse Spectrum Gaussian Processes [15] (SSGPs), which can be learned from real data.”
Improving Multi-Step Prediction of Learned Time Series Models (2015, background): “Recent work in kernel machines enables us to instead approximate φ through generating random Fourier features (Rahimi and Recht 2007; 2008; Le, Sarlós, and Smola 2013; Lázaro-Gredilla 2010).”
A la Carte - Learning Fast Kernels (2014, background): “SSGPR Sparse Spectrum Gaussian Process Regression is a kitchen sinks (RKS) based model which individually optimizes the locations of all spectral frequencies [L´azaro-Gredilla et al., 2010].”
Automatic Construction and Natural-Language Description of Nonparametric Regression Models (2014, methodology): “Kernel Learning Sparse spectrum GPs (Lázaro-Gredilla et al., 2010) approximate the spectral density of a stationary kernel function using delta functions; this corresponds to kernels of the form ∑ cos.”
Thoughts on Massively Scalable Gaussian Processes (2015, methodology): “GP inference and learning can be performed in O ( n ) using KISS-GP, a signiﬁcant advance over the more standard O ( m 2 n ) scaling of fast GP methods (Qui˜nonero-Candela and Ras-mussen, 2005a; L´azaro-Gredilla et al., 2010).”
Automated Machine Learning on Big Data using Stochastic Algorithm Tuning (2014, methodology): “[18] provide an alternative method to select the eigenpairs by Monte-Carlo sampling of the kernel function’s spectral density.”
Synthesizing Neural Network Controllers with Probabilistic Model-Based Reinforcement Learning (2018, methodology): “(4a) illustrates the difference in performace between PILCO using sparse spectrum GP (SSGP) regression [32] for the dynamics and two versions of Deep-PILCO using BNN dynamics: one using binary dropout with dropout probability p = 0.”
Automatic model construction with Gaussian processes (2014, methodology): “5.1 Relating deep neural networks to deep GPs . . . . . . . . . . . . . . . .”
Functional Variational Bayesian Neural Networks (2019, methodology): “However, sparse GP methods (Lázaro-Gredilla et al., 2010; Snelson & Ghahramani, 2006; Titsias, 2009; Hensman et al., 2013; 2015; Krauth et al., 2016) still suffer for very large dataset, while random feature methods (Rahimi & Recht, 2008; Le et al., 2013) and KISS-GP (Wilson & Nickisch, 2015;…”
Blitzkriging: Kronecker-structured Stochastic Gaussian Processes (2015, unknown): “[10].”
Random Feature Expansions for Deep Gaussian Processes (2016, methodology): “Random Fourier features for large scale kernel machines were proposed in Rahimi & Recht (2008), and their application to GP s appears in L´azaro-Gredilla et al. (2010).”
When Gaussian Process Meets Big Data: A Review of Scalable GPs (2018, background): “Alternatively, the sparse spectrum GP (SSGP) [61] and its variational variants [62]–[64] elegantly address this issue by reconstructing the Bayesian linear model from spectral representation (Fourier features), resulting in the stationary kernel where s r ∈ R d represents the spectral frequencies.”
Dropout as a Bayesian Approximation : Insights and Applications (2015, methodology): “As was discussed in (Lázaro-Gredilla et al., 2010) this leads to model overfitting .”
Learning-Based Model Predictive Control: Toward Safe Learning in Control (2020, background): “, via inducing points (93) or selected basis functions (94)], and neglecting or simplifying the variance dynamics (i.”
Cautious Model Predictive Control Using Gaussian Process Regression (2017, methodology): “Various approximation techniques have been proposed to reduce the computational complexity of GPs, for instance sparse spectrum approximation [12], which was shown in the context of MPC in [6].”