ML Anthology

SampleRank: Training Factor Graphs with Atomic Gradients

Michael L. Wick, Khashayar Rohanimanesh, Kedar Bellare, Aron Culotta, Andrew McCallum
ICML 2011 pp. 777-784
/icml/2011/wick2011icml-samplerank/

Abstract

We present SampleRank, an alternative to contrastive divergence (CD) for estimating parameters in complex graphical models. SampleRank harnesses a user-provided loss function to distribute stochastic gradients across an MCMC chain. As a result, parameter updates can be computed between arbitrary MCMC states. SampleRank is not only faster than CD, but also achieves better accuracy in practice (up to 23\% error reduction on noun-phrase coreference).

PDF Semantic Scholar

Cite

Text

Wick et al. "SampleRank: Training Factor Graphs with Atomic Gradients." International Conference on Machine Learning, 2011.

Markdown

[Wick et al. "SampleRank: Training Factor Graphs with Atomic Gradients." International Conference on Machine Learning, 2011.](https://mlanthology.org/icml/2011/wick2011icml-samplerank/)

BibTeX

@inproceedings{wick2011icml-samplerank,
  title     = {{SampleRank: Training Factor Graphs with Atomic Gradients}},
  author    = {Wick, Michael L. and Rohanimanesh, Khashayar and Bellare, Kedar and Culotta, Aron and McCallum, Andrew},
  booktitle = {International Conference on Machine Learning},
  year      = {2011},
  pages     = {777-784},
  url       = {https://mlanthology.org/icml/2011/wick2011icml-samplerank/}
}