Multi-task Regularization of Generative Similarity Models

Home >> Publications >> SIMBAD 2011 - Multi-task SDA Projects    Blog    Bio
Luca Cazzanti, Sergey Feldman, Maya R. Gupta, and Michael Gabbay
PDF | Video and slides of the talk

Abstract - We investigate a multi-task approach to similarity discriminant analysis, where we propose treating the estimation of the different class-conditional distributions of the pairwise similarities as multiple tasks. We show that regularizing these estimates together using a least-squares regularization weighted by a task-relatedness matrix can reduce the resulting maximum a posteriori classification errors. Results are given for benchmark data sets spanning a range of applications. In addition, we present a new application of similarity-based learning to analyzing the rhetoric of multiple insurgent groups in Iraq. We show how to produce the necessary task relatedness information from standard given training data, as well as how to derive task-relatedness information if given side information about the class relatedness.

Datasets - AuralSonar | AmazonBinary | Patrol | Voting | Protein | Face Recognition | Insurgent Rhetoric
Software - Matlab implementation of multi-task local SDA regularization, plus all other flavors of local SDA. Visit the downloads page for more software.

BibTex -
TITLE = {Multi-task Regularization of Generative Similarity Models},
AUTHOR = {L. Cazzanti and S. Feldman and M. R. Gupta and M. Gabbay},
BOOKTITLE = {Proc. Intl. Workshop on Similarity-based Pattern Recognition ({SIMBAD 11})},
EDITOR = {M. Pelillo and E. R. Hancock},
VOLUME = {LNCS 7005},
ADDRESS = {Venice, Italy},
MONTH = {September},
YEAR = {2011},
PUBLISHER = {Springer}}