A Stick-Breaking Likelihood for Categorical Data Analysis with Latent Gaussian Models
نویسندگان
چکیده
The development of accurate models and efficient algorithms for the analysis of multivariate categorical data are important and longstanding problems in machine learning and computational statistics. In this paper, we focus on modeling categorical data using Latent Gaussian Models (LGMs). We propose a novel stick-breaking likelihood function for categorical LGMs that exploits accurate linear and quadratic bounds on the logistic log-partition function, leading to an effective variational inference and learning framework. We thoroughly compare our approach to existing algorithms for multinomial logit/probit likelihoods on several problems, including inference in multinomial Gaussian process classification and learning in latent factor models. Our extensive comparisons demonstrate that our stick-breaking model effectively captures correlation in discrete data and is well suited for the analysis of categorical data.
منابع مشابه
A Stick-Breaking Likelihood for Categorical Data Analysis with Latent Gaussian Models
The development of accurate models and efficient algorithms for the analysis of multivariate categorical data are important and longstanding problems in machine learning and computational statistics. In this paper, we focus on modeling categorical data using Latent Gaussian Models (LGMs). We propose a novel logistic stick-breaking likelihood function for categorical LGMs that can exploit recent...
متن کاملParameter Estimation in Spatial Generalized Linear Mixed Models with Skew Gaussian Random Effects using Laplace Approximation
Spatial generalized linear mixed models are used commonly for modelling non-Gaussian discrete spatial responses. We present an algorithm for parameter estimation of the models using Laplace approximation of likelihood function. In these models, the spatial correlation structure of data is carried out by random effects or latent variables. In most spatial analysis, it is assumed that rando...
متن کاملAn application of Measurement error evaluation using latent class analysis
Latent class analysis (LCA) is a method of evaluating non sampling errors, especially measurement error in categorical data. Biemer (2011) introduced four latent class modeling approaches: probability model parameterization, log linear model, modified path model, and graphical model using path diagrams. These models are interchangeable. Latent class probability models express l...
متن کاملDependent Multinomial Models Made Easy: Stick-Breaking with the Polya-gamma Augmentation
Many practical modeling problems involve discrete data that are best represented as draws from multinomial or categorical distributions. For example, nucleotides in a DNA sequence, children’s names in a given state and year, and text documents are all commonly modeled with multinomial distributions. In all of these cases, we expect some form of dependency between the draws: the nucleotide at on...
متن کاملEstimation of latent Gaussian ARMA models for categorical behaviour data
We consider the tting of latent Gaussian models to categorical time series of cow feeding data. We derive a spectral quasi-likelihood for the data, and compare it with least squares ts to autocorrelations and MCMC estimators of the parameters in thresholded ARMA processes. We show that the spectral method is more e cient than least squares and far faster than MCMC.
متن کامل