Infinite latent feature models and the Indian buffet process

By T. Griffiths, Zoubin Ghahramani. 2005

860 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

Federated Learning with Matched Averaging (2020, methodology): “…model based on the Beta-Bernoulli process (BBP) (Thibaux & Jordan, 2007), where similarity c(wjl, θi) is the corresponding posterior probability of jth client neuron l generated from a Gaussian with mean θi, and and f(·) are guided by the Indian Buffet Process prior (Ghahramani & Griffiths, 2005).”
Unsupervised Semantic Parsing of Video Collections (2015, methodology): “In the IBP analogy, each video is a customer and each activity step is a dish in the buffet.”
Stick-Breaking Variational Autoencoders (2016, methodology): “The best known model of this kind is the inﬁnite binary latent feature model deﬁned by the Indian Buffet Process (Ghahramani & Grifﬁths, 2005); but its discrete latent variables prevent this model from admitting fully differentiable inference.”
Exploiting compositionality to explore a large space of model structures (2012, methodology): “To apply the rule G → BG + G or G → GB + G, we perform accelerated collapsed Gibbs sampling (Doshi-Velez and Ghahramani, 2009) over the binary variables in a linear-Gaussian Indian Buffet Process (Griffiths and Ghahramani, 2005)”
Distributed large-scale natural graph factorization (2013, background): “There exists a wide range of objective functions for graph reconstruction and factorization [13].”
Sharing Features among Dynamical Systems with Beta Processes (2009, methodology): “In evaluating this expression, we have exploited the exchangeability of the IBP [9], which follows directly f rom the beta process construction [22].”
Bayesian Nonparametric Weight Factorization for Continual Learning (2020, methodology): “the Indian Buffet Process (IBP) [9] as a prior for task-specific factor selection: r t,k ∼ N (0, 1) (5) v t,i ∼ Beta(α, 1) (6) π t,k = k ∏”
Learning the Structure of Deep Convolutional Networks (2015, methodology): “Fortunately, model structure learning has been elegantly explored by several generative model priors, such as the Indian Buffet Process (IBP) [9, 1] or Beta Process (BP) [21].”
Variational Inference for the Indian Buffet Process (2009, methodology): “These approaches include Gibbs sampling (Griffiths and Ghahramani, 2005), which may be augmented with Metropolis split-merge proposals (Meeds et al.”
Latent feature learning in social media network (2013, background): “We describe this prior as: ZA ~ IBP(a) (6) ZB ~ IBP(ß) (7) where a and ß control the growth rate of the new features.”
Infinite multiple membership relational modeling for complex networks (2011, unknown): “We learn the effective number of features through the Indian buffet process (IBP) representation [12], which defines a distribution over unbounded binary matrices,”
Beta-Negative Binomial Process and Poisson Factor Analysis (2011, background): “Unlike the IBP, however, the msIBP predictive posterior distribution of X n +1”
Generative Modeling of Infinite Occluded Objects for Compositional Scene Representation (2019, methodology): “…be seen as a binary matrix with one row and infinite columns which is obtained using the stick-breaking construction (Teh et al., 2007) for the Indian Buffet Process (IBP) (Ghahramani & Griffiths, 2006; Griffiths & Ghahramani, 2011). zdep is a binary matrix with finite rows and infinite columns.”
Learning the Structure of Deep Sparse Graphical Models (2009, methodology): “We can use the Indian buffet process (IBP) [Griffiths and Ghahramani, 2006] to allow this matrix to have an infinite number of columns.”