Divide-and-Conquer Learning by Anchoring a Conical Hull
Tianyi Zhou, Jeffrey A. Bilmes, Carlos Guestrin

Citation
Tianyi Zhou, Jeffrey A. Bilmes, Carlos Guestrin. "Divide-and-Conquer Learning by Anchoring a Conical Hull". Talk or presentation, October, 2014; Poster presented at the 2014 TerraSwarm Annual Meeting.

Abstract
We propose a divide-and-conquer learning scheme for several core machine learning problems, and bring hundreds to thousands times of speedup as well as more interpretable output model. The proposed method 1) Makes big data challenge tractable: Terraswarm datasets could be very large. Our method only needs to select a small subset of real data instances by fast divide-and-conquer algorithm, with non-iterative procedure on extremely low-dimensional projections. 2) Has Broad applications: feature/topic extraction and clustering on NLP/vision/speech data. 3) Makes learning more explainable: select a small subset of real data instances (anchors) to build the parameters of many statistical learning models.

Electronic downloads

Internal. This publication has been marked by the author for TerraSwarm-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Tianyi Zhou, Jeffrey A. Bilmes, Carlos Guestrin. <a
  href="http://www.terraswarm.org/pubs/439.html"><i>Divide-and-Conquer
  Learning by Anchoring a Conical Hull</i></a>,
  Talk or presentation,  October, 2014; Poster presented at
  the <a
  href="http://www.terraswarm.org/conferences/14/annual"
  >2014 TerraSwarm Annual Meeting</a>.

• Plain text

  Tianyi Zhou, Jeffrey A. Bilmes, Carlos Guestrin.
  "Divide-and-Conquer Learning by Anchoring a Conical
  Hull". Talk or presentation,  October, 2014; Poster
  presented at the <a
  href="http://www.terraswarm.org/conferences/14/annual"
  >2014 TerraSwarm Annual Meeting</a>.

• BibTeX

  @presentation{ZhouBilmesGuestrin14_DivideandConquerLearningByAnchoringConicalHull,
      author = {Tianyi Zhou and Jeffrey A. Bilmes and Carlos
                Guestrin},
      title = {Divide-and-Conquer Learning by Anchoring a Conical
                Hull},
      month = {October},
      year = {2014},
      note = {Poster presented at the <a
                href="http://www.terraswarm.org/conferences/14/annual"
                >2014 TerraSwarm Annual Meeting</a>.},
      abstract = {We propose a divide-and-conquer learning scheme
                for several core machine learning problems, and
                bring hundreds to thousands times of speedup as
                well as more interpretable output model. The
                proposed method 1) Makes big data challenge
                tractable: Terraswarm datasets could be very
                large. Our method only needs to select a small
                subset of real data instances by fast
                divide-and-conquer algorithm, with non-iterative
                procedure on extremely low-dimensional
                projections. 2) Has Broad applications:
                feature/topic extraction and clustering on
                NLP/vision/speech data. 3) Makes learning more
                explainable: select a small subset of real data
                instances (anchors) to build the parameters of
                many statistical learning models. },
      URL = {http://terraswarm.org/pubs/439.html}
  }
  

Posted by Tianyi Zhou on 4 Nov 2014.
Groups: services