Training conditional random fields for maximum labelwise accuracy

Samuel S. Gross; Olga Russakovsky; Chuong B. Do; Serafim Batzoglou

Training conditional random fields for maximum labelwise accuracy

Samuel S. Gross, Olga Russakovsky, Chuong B. Do, Serafim Batzoglou

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We consider the problem of training a conditional random field (CRF) to maximize per-label predictive accuracy on a training set, an approach motivated by the principle of empirical risk minimization. We give a gradient-based procedure for minimizing an arbitrarily accurate approximation of the empirical risk under a Hamming loss function. In experiments with both simulated and real data, our optimization procedure gives significantly better testing performance than several current approaches for CRF training, especially in situations of high label noise.

Original language	English (US)
Title of host publication	Advances in Neural Information Processing Systems 19 - Proceedings of the 2006 Conference
Pages	529-536
Number of pages	8
State	Published - 2007
Externally published	Yes
Event	20th Annual Conference on Neural Information Processing Systems, NIPS 2006 - Vancouver, BC, Canada Duration: Dec 4 2006 → Dec 7 2006

Publication series

Name	Advances in Neural Information Processing Systems
ISSN (Print)	1049-5258

Other

Other	20th Annual Conference on Neural Information Processing Systems, NIPS 2006
Country/Territory	Canada
City	Vancouver, BC
Period	12/4/06 → 12/7/06

All Science Journal Classification (ASJC) codes

Computer Networks and Communications
Information Systems
Signal Processing

Cite this

@inproceedings{8b5027db5f124d509e82e5c2badeea21,

title = "Training conditional random fields for maximum labelwise accuracy",

abstract = "We consider the problem of training a conditional random field (CRF) to maximize per-label predictive accuracy on a training set, an approach motivated by the principle of empirical risk minimization. We give a gradient-based procedure for minimizing an arbitrarily accurate approximation of the empirical risk under a Hamming loss function. In experiments with both simulated and real data, our optimization procedure gives significantly better testing performance than several current approaches for CRF training, especially in situations of high label noise.",

author = "Gross, {Samuel S.} and Olga Russakovsky and Do, {Chuong B.} and Serafim Batzoglou",

year = "2007",

language = "English (US)",

isbn = "9780262195683",

series = "Advances in Neural Information Processing Systems",

pages = "529--536",

booktitle = "Advances in Neural Information Processing Systems 19 - Proceedings of the 2006 Conference",

note = "20th Annual Conference on Neural Information Processing Systems, NIPS 2006 ; Conference date: 04-12-2006 Through 07-12-2006",

}

Gross, SS, Russakovsky, O, Do, CB & Batzoglou, S 2007, Training conditional random fields for maximum labelwise accuracy. in Advances in Neural Information Processing Systems 19 - Proceedings of the 2006 Conference. Advances in Neural Information Processing Systems, pp. 529-536, 20th Annual Conference on Neural Information Processing Systems, NIPS 2006, Vancouver, BC, Canada, 12/4/06.

Training conditional random fields for maximum labelwise accuracy. / Gross, Samuel S.; Russakovsky, Olga; Do, Chuong B. et al.
Advances in Neural Information Processing Systems 19 - Proceedings of the 2006 Conference. 2007. p. 529-536 (Advances in Neural Information Processing Systems).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Training conditional random fields for maximum labelwise accuracy

AU - Gross, Samuel S.

AU - Russakovsky, Olga

AU - Do, Chuong B.

AU - Batzoglou, Serafim

PY - 2007

Y1 - 2007

N2 - We consider the problem of training a conditional random field (CRF) to maximize per-label predictive accuracy on a training set, an approach motivated by the principle of empirical risk minimization. We give a gradient-based procedure for minimizing an arbitrarily accurate approximation of the empirical risk under a Hamming loss function. In experiments with both simulated and real data, our optimization procedure gives significantly better testing performance than several current approaches for CRF training, especially in situations of high label noise.

AB - We consider the problem of training a conditional random field (CRF) to maximize per-label predictive accuracy on a training set, an approach motivated by the principle of empirical risk minimization. We give a gradient-based procedure for minimizing an arbitrarily accurate approximation of the empirical risk under a Hamming loss function. In experiments with both simulated and real data, our optimization procedure gives significantly better testing performance than several current approaches for CRF training, especially in situations of high label noise.

UR - http://www.scopus.com/inward/record.url?scp=34548413443&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34548413443&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:34548413443

SN - 9780262195683

T3 - Advances in Neural Information Processing Systems

SP - 529

EP - 536

BT - Advances in Neural Information Processing Systems 19 - Proceedings of the 2006 Conference

T2 - 20th Annual Conference on Neural Information Processing Systems, NIPS 2006

Y2 - 4 December 2006 through 7 December 2006

ER -

Training conditional random fields for maximum labelwise accuracy

Abstract

Publication series

Other

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this