University of Illinois Urbana-Champaign

Distributed Strategies for Topic Modeling

Jindal, Prateek; Hockenmaier, Julia; Kale, Laxmikant V.

Permalink

https://hdl.handle.net/2142/46405

Description

Title
Distributed Strategies for Topic Modeling
Author(s)
  • Jindal, Prateek
  • Hockenmaier, Julia
  • Kale, Laxmikant V.
Issue Date
2013-12-04
Keyword(s)
  • Topic Models
  • Latent Dirichlet Allocation
  • Parallel Computing
  • Charm
Abstract
Topic modeling algorithms (like Latent Dirichlet Allocation) tend to be very slow when run over large document collections. In this presentation, we discuss distributed strategies for topic modeling. We use Charm++ as our parallelization framework. Our results show that parallelization can considerably increase the efficiency of topic modeling.
Type of Resource
text
Language
en
Permalink
http://hdl.handle.net/2142/46405
Copyright and License Information
2013 by Prateek Jindal

Owning Collections

Research and Tech Reports - Computer Science PRIMARY