Sparse Partially Collapsed MCMC for Parallel Inference in Topic Models

TL;DR

This paper introduces a parallel sparse partially collapsed Gibbs sampler for LDA that improves inference speed and efficiency, especially on large corpora, while maintaining accuracy.

Contribution

It proposes a novel parallel sparse partially collapsed Gibbs sampler for topic models, demonstrating improved speed and scalability over existing methods.

Findings

Partial collapsing causes less statistical inefficiency than assumed.

Parallelization and sparsity significantly speed up inference on large datasets.

The method scales well with corpus size, maintaining accuracy and efficiency.

Abstract

Topic models, and more specifically the class of Latent Dirichlet Allocation (LDA), are widely used for probabilistic modeling of text. MCMC sampling from the posterior distribution is typically performed using a collapsed Gibbs sampler. We propose a parallel sparse partially collapsed Gibbs sampler and compare its speed and efficiency to state-of-the-art samplers for topic models on five well-known text corpora of differing sizes and properties. In particular, we propose and compare two different strategies for sampling the parameter block with latent topic indicators. The experiments show that the increase in statistical inefficiency from only partial collapsing is smaller than commonly assumed, and can be more than compensated by the speedup from parallelization and sparsity on larger corpora. We also prove that the partially collapsed samplers scale well with the size of the corpus. The proposed algorithm is fast, efficient, exact, and can be used in more modeling situations than the ordinary collapsed sampler.