Variational bounds for mixed-data factor analysis

By Mohammad Emtiyaz Khan, Benjamin M. Marlin, Guillaume Bouchard, Kevin P. Murphy. 2010

66 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

Variational Inference: A Review for Statisticians (2016, background): “Blei and Lafferty (2007) later adapted their idea to nonconjugate topic models, and researchers have continued to improve the original bound (Khan et al. 2010; Marlin, Khan, and Murphy 2011; Ermis and Bouchard 2014).”
Incremental Majorization-Minimization Optimization with Application to Large-Scale Machine Learning (2014, background): “Such surrogates appear frequently in the statistics and machine learning literature [7, 31, 33].”
Monte Carlo Gradient Estimation in Machine Learning (2019, methodology): “Local variational inference methods developed several bounding methods, such as Bohning and Jaakola’s bounds (Khan et al., 2010), and piecewise bounds (Marlin et al.”
Affine Independent Variational Inference (2012, methodology): “For other models, for instance mixed data factor analysis [15], different distributions for blocks of variables of {vd}Dd=1 may be optimal.”
Piecewise Bounds for Estimating Bernoulli-Logistic Latent Gaussian Models (2011, methodology): “Latent Gaussian Models (LGMs) are an important class of probabilistic models that includes factor analysis and probabilistic principal components analysis for continuous data (Tipping & Bishop, 1999), as well as binary and multinomial factor analysis for discrete data (Wedel & Kamakura, 2001; Collin”
Infinite Latent SVM for Classification and Multi-task Learning (2011, background): “Moreover, expectation can be more robust than taking the mode [18], and it has been used in [31, 32].”
1-Bit Matrix Completion (2012, background): “There are a number of compelling algorithmic papers on these subjects, including [19, 22, 38, 57, 64], which suggest positive answers on simulated and real-world data.”
Iterative Splits of Quadratic Bounds for Scalable Binary Tensor Factorization (2014, methodology): “Khan et al [13] used Jaakkola’s bound for binary observations and Bohning’s bound for multinomial observations [1].”
Variational inference in nonconjugate models (2012, unknown): “…have developed a variety of strategies for a variety of models, including approximations (Braun and McAuliffe, 2010; Ahmed and Xing, 2007), alternative bounds (Jaakkola and Jordan, 1997; Blei and Lafferty, 2006, 2007; Khan et al., 2010), and numerical quadrature (Honkela and Valpola, 2004).”
Non-conjugate Variational Message Passing for Multinomial and Binary Regression (2011, background): “The second line of work applying VB to non-conjugate models involves deriving lower bounds to approximate the expectations [9, 18, 5, 10, 11] required to calculate the KL divergence.”
General Latent Feature Models for Heterogeneous Datasets (2017, methodology): “Note that test loglikelihood results are provided for the baseline MDFA, since the exact computation of the likelihood is not tractable and (Khan et al., 2010) only provide a bound on the likelihood.”
Variational Approximations for Generalized Linear Latent Variable Models (2017, unknown): “Variational EM algorithms have also been proposed specifically for random e ff ects item response theory models (Rijmen and Jeon, 2013) and factor analysis (Khan et al., 2010), but none so far have considered the broader GLLVM framework.”
Nonparametric variational inference (2012, background): “Several researchers use specialized approximations for certain classes of models, such as those with logistic nonlinearities (e.g., Jaakkola & Jordan, 2000; Khan et al., 2010).”
Scalable Variational Inference for Bayesian Variable Selection in Regression, and Its Accuracy in Genetic Association Studies (2012, unknown): “Motivated by this, others (e.g. Blei et al. 2003; Khan et al. 2010) have proposed to replace the intractable maximum likelihood estimator for θ with a θ that maximizes the best lower bound, F (θ;φ(θ)).”