conftrace_

2014 NeurIPS NeurIPS 2014

Parallel Sampling of HDPs using Sub-Cluster Splits

Abstract

We develop a sampling technique for Hierarchical Dirichlet process models. The parallel algorithm builds upon [Chang & Fisher 2013] by proposing large split and merge moves based on learned sub-clusters. The additional global split and merge moves drastically improve convergence in the experimental results. Furthermore, we discover that cross-validation techniques do not adequately determine convergence, and that previous sampling methods converge slower than were previously expected.

🐣 Hot Topic Early Bird - latent variable model

Authors

Jason Chang, John W. Fisher III

Topics

Artificial Intelligence > Bayesian & Probabilistic > Bayesian Learning Machine Learning > Core Methods > Clustering Machine Learning > Learning Types > Unsupervised Learning Machine Learning > Optimization & Theory > Bayesian Inference Machine Learning > Optimization & Theory > Stochastic Processes Machine Learning > Bayesian & Probabilistic > Probabilistic Modeling Artificial Intelligence > Bayesian & Probabilistic > Bayesian Inference Machine Learning > Bayesian & Probabilistic > Markov Chain Monte Carlo Machine Learning > Bayesian & Probabilistic > Nonparametric Bayesian

Keywords

bayesian nonparametrics probabilistic modeling topic modeling markov chain monte carlo gibbs sampling hierarchical dirichlet process mcmc sampling bayesian nonparametric model latent variable model split-merge algorithm parallel sampling split-merge sampling split-merge move

More from NeurIPS 2014

All NeurIPS 2014 papers →

Constrained convex minimization via model-based excessive gap

Metric Learning for Temporal Sequence Alignment

An Accelerated Proximal Coordinate Gradient Method

Universal Option Models

Convolutional Kernel Networks

Coresets for k-Segmentation of Streaming Data