Variational Inference for the Indian Buffet Process

By Finale Doshi-Velez, Kurt T. Miller, Jurgen Van Gael, Yee Whye Teh. 2009

158 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

Bayesian Structure Adaptation for Continual Learning (2019, methodology): “…infer the number of nodes in each hidden layer, we use the IBP prior (Grifﬁths & Ghahramani, 2011), whose truncated stick-breaking process (Doshi et al., 2009) construction for each entry of B is as follows for d ∈ 1 , ..., D , where D denotes the number of input nodes for this hidden…”
Multicanonical Stochastic Variational Inference (2014, methodology): “Details on the inference updates can be found in [3].”
Nonparametric bayesian multitask collaborative filtering (2013, methodology): “Variational Bayesian inference for the IBP is based on an alternative formulation of p(Z) [7], namely the stick-breaking construction of the IBP [31]: It has been shown that the prior (1) obtained by the IBP can be equivalently expressed under the following hierarchical Bayesian construction”
Flexible Modeling of Latent Task Structures in Multitask Learning (2012, methodology): “As this model is infinite and combinatorial in nature, exact inference is intractable and sampling-based inference may take too long to converge (Doshi-Velez et al., 2009; Blei & Jordan, 2006).”
Online Variational Inference for the Hierarchical Dirichlet Process (2011, methodology): “For example, the Indian Buffet process (IBP) [25, 26, 27] might be another model that can use an efficient online variational inference algorithm for large and streaming data sets.”
Infinite Latent SVM for Classification and Multi-task Learning (2011, methodology): “For p(ν), we have the same update rules as in [9].”
Spectral Methods for Indian Buffet Process Inference (2014, background): “Moreover, we show that empirically the spectral algorithm provides higher accuracy and lower runtime than variational methods [14].”
Variational Tempering (2014, methodology): “We also carried out experiments on the Factorial Mixture Model (FMM) (Ghahra-mani, 1995; Doshi et al., 2009).”
Indian Buffet Neural Networks for Continual Learning (2019, methodology): “Catastrophic forgetting is overcome by repeated application of the Bayesian update rule, embedded within variational inference [12, 2].”
Variational Russian Roulette for Deep Bayesian Nonparametrics (2019, methodology): “Previous methods for variational inference in such models rely on a truncated approximation, which places an upper bound on the size of the latent space under the approximate posterior (Blei & Jordan, 2004; Doshi-Velez et al., 2009).”
A Coupled Indian Buffet Process Model for Collaborative Filtering (2012, methodology): “Regarding the stick-breaking variables of the employed IBPs, the posterior distributions are similar to the one derived in Doshi-Velez et al. (2009).”
Generative Modeling of Infinite Occluded Objects for Compositional Scene Representation (2019, methodology): “The second is to apply the multinomial approximation to Eν [log(1− ∏k k′=1 νk′)] as mentioned in (Doshi et al., 2009) and obtain an upper bound of DKL(qφ(zindk |sstn·k )||p(zindk |ν1:k)) that has a closedform solution.”
General Latent Feature Models for Heterogeneous Datasets (2017, background): “The standard linear-Gaussian IBP model assumes binary latent features, Gaussian weighting factors, that capture the influence of every latent feature in each observed attribute, and observations, leading to a conjugate model that permits the use of fast inference algorithms (Doshi-Velez and Ghahrama”
Discovery of Treatments from Text Corpora (2016, methodology): “A standard derivation that builds on Doshi-Velez et al. (2009) leads to the following distributional forms and update steps:”