More data means less inference: A pseudo-max approach to structured learning
Name
Jaakkola_More data.pdf
Size
364.13 KB
Format
Adobe PDF
Checksum (MD5)
4ba5e4325e20d95c73130f578259fc76
Author(s)
Sontag, David
Meshi, Ofer
Jaakkola, Tommi S.
Globerson, Amir
Date Issued
December 2010
Journal
Poster Session paper of the Twenty-Fourth Annual Conference on Neural Information Processing Systems, NIPS 2010
Publisher
Neural Information Processing Systems Foundation
Citation
Sontag, David et al. "More data means less inference: A pseudo-max approach to structured learning." in Proceedins of the Twenty-Fourth Annual Conference on Neural Information Processing Systems, NIPS 2010, Poster Session, December 6-9, Vancouver, B.C., Canada.
Version
Author's final manuscript
Abstract
The problem of learning to predict structured labels is of key importance in many applications. However, for general graph structure both learning and inference in this setting are intractable. Here we show that it is possible to circumvent this difficulty when the input distribution is rich enough via a method similar in spirit to pseudo-likelihood. We show how our new method achieves consistency, and illustrate empirically that it indeed performs as well as exact methods when sufficiently large training sets are used.
MIT Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike 3.0
Persistent DSpace Link
DOI of Published Version
http://books.nips.cc/papers/files/nips23/NIPS2010_0809.pdf
