Discriminative learning via semidefinite probabilistic models

Koby Crammer*, Amir Globerson

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Discriminative linear models are a popular tool in machine learning. These can be generally divided into two types: linear classifiers, such as support vector machines (SVMs), which are well studied and provide state-of-the-art results, and probabilistic models such as logistic regression. One shortcoming of SVMs is that their output (known as the "margin") is not calibrated, so that it is difficult to incorporate such models as components of larger systems. This problem is solved in the probabilistic approach. We combine these two approaches above by constructing a model which is both linear in the model parameters and probabilistic, thus allowing maximum margin training with calibrated outputs. Our model assumes that classes correspond to linear sub-spaces (rather than to half spaces), a view which is closely related to concepts in quantum detection theory. The corresponding optimization problems are semidefinite programs which can be solved efficiently. We illustrate the performance of our algorithm on real world datasets, and show that it outperforms second-order kernel methods.

Original languageEnglish
Title of host publicationProceedings of the 22nd Conference on Uncertainty in Artificial Intelligence, UAI 2006
Pages98-105
Number of pages8
StatePublished - 2006
Externally publishedYes
Event22nd Conference on Uncertainty in Artificial Intelligence, UAI 2006 - Cambridge, MA, United States
Duration: 13 Jul 200616 Jul 2006

Publication series

NameProceedings of the 22nd Conference on Uncertainty in Artificial Intelligence, UAI 2006

Conference

Conference22nd Conference on Uncertainty in Artificial Intelligence, UAI 2006
Country/TerritoryUnited States
CityCambridge, MA
Period13/07/0616/07/06

Fingerprint

Dive into the research topics of 'Discriminative learning via semidefinite probabilistic models'. Together they form a unique fingerprint.

Cite this