Scalable Inference for Latent Dirichlet Allocation

TL;DR

This paper presents a scalable, asynchronous distributed inference method for Latent Dirichlet Allocation that balances speed and accuracy, suitable for heterogeneous computing clusters.

Contribution

It introduces a simple, tunable approximation method for distributed LDA inference that is asynchronous and adaptable to different hardware environments.

Findings

Efficient distributed inference with adjustable accuracy

Asynchronous approach suitable for heterogeneous clusters

Demonstrates scalability and flexibility in LDA learning

Abstract

We investigate the problem of learning a topic model - the well-known Latent Dirichlet Allocation - in a distributed manner, using a cluster of C processors and dividing the corpus to be learned equally among them. We propose a simple approximated method that can be tuned, trading speed for accuracy according to the task at hand. Our approach is asynchronous, and therefore suitable for clusters of heterogenous machines.