Sparse Greedy Gaussian Process Regression

By Alex Smola, Peter L. Bartlett. 2000

442 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Kernel Interpolation for Scalable Structured Gaussian Processes (KISS-GP) (2015, neutral): “Turner (2010) and Cunningham et al.”
Building Support Vector Machines with Reduced Classifier Complexity (2006, background): “Such an approach has been popular in Gaussian processes (Smola and Bartlett, 2001; Seeger et al., 2003; Keerthi and Chu, 2006).”
A Practical and Conceptual Framework for Learning in Control (2010, methodology): “26), are functionally dependent on both the policy parameter vector ψ and the moments μt−1 and Σt−1 of the state distribution p(xt−1) at time 12The concept of the pseudo-training set is closely related to the ideas of inducing inputs used in the sparse GP approximation by Snelson and Ghahramani (200”
Fast Forward Selection to Speed Up Sparse Gaussian Process Regression (2003, methodology): “Smola and Bartlett (2001) and Luo and Wahba (1997) use greedy forward selection schemes just as we do, although our framework is somewhat different.”
Kernel Matching Pursuit (2002, background): “Another recent work, very much related to ours, that also uses a Matching-Pursuit like algorithm is (Smola and Bartlett, 2001).”
Gaussian processes:iterative sparse approximations (2002, methodology): “36) has also been used for regression [Smola and Schölkopf 2000; Smola and Schölkopf 2001].”
Constrained Bayesian Optimization for Automatic Chemical Design (2017, methodology): “ational Free Energy methods [121] and methods based on Expectation Propagation (EP) [110,91,11]. From the methods provided above, the two most popular are FITC and VFE [3], as unlike previous methods [108, 102, 92], they provide an approximation to the marginal likelihood which may be leveraged in c”
Bayes Meets Bellman: The Gaussian Process Approach to Temporal Difference Learning (2003, background): “¢(x~) = Ea~,j¢(:~j) + ¢~es, where IlCresH 2 v. j-----1 (3.15) In order to be able to compute at at each time step, we need to update I~t i whenever the dictionary is appended with a new state.”
Approximation Methods for Gaussian Process Regression (2007, methodology): “One way of obtaining an equivalent (5)Wary of this fact, Smola and Bartlett (2001) propose using the predictive variances of the SoD method, or a more accurate but computationally costly alternative (more details are given in Quiñonero-Candela, 2004, Chapter 3).”
Computationally Efficient Convolved Multiple Output Gaussian Processes (2011, methodology): “Recently, Titsias (2009) highlighted how optimizing inducing variables can be problematic as they introduce many hyperparameters in the likelihood term.”
PAC-Bayesian Generalisation Error Bounds for Gaussian Process Classification (2003, unknown): “If we do a sparse greedy maximization of q, allowing only a controlled number k of components of u to become nonzero, we can compute upper and lower bounds on σ2 ∗ in O(nk 2) (see Smola and Bartlett (2001)), without having to know more than k rows of A.”
Learning for Larger Datasets with the Gaussian Process Latent Variable Model (2007, background): “A large body of recent work has been focussed on approximating the covariance function with a low rank approximation [Smola and Bartlett, 2001, Williams and Seeger, 2001, Tresp, 2000, Schwaighofer and Tresp, 2003, Csató and Opper, 2002, Seeger et al., 2003].”
Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning) (2005, unknown): “Sollich [1999] derives a number of more accurate approximations to the learning curve than eq.”
Information Theory, Inference, and Learning Algorithms (2004, background): “Methods for reducing the complexity of data modelling with Gaussian processes remain an active research area (Poggio and Girosi, 1990; Luo and Wahba, 1997; Tresp, 2000; Williams and Seeger, 2001; Smola and Bartlett, 2001; Rasmussen, 2002; Seeger et al., 2003; Opper and Winther, 2000).”
A Unifying View of Sparse Approximate Gaussian Process Regression (2005, methodology): “It has been proposed to maximize the effective posterior instead of the effective marginal likelihood (Smola and Bartlett, 2001).”