PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

The Entire Quantile Path of a Risk-Agnostic SVM Classifier
Jin Yu, Vishwanathan S.V.N. and Jian Zhang
Conference on Uncertainty in Artificial Intelligence 2009.

Abstract

A quantile binary classifier uses the rule: Classify x as +1 if P(Y=1|X=x) >= tau, and as -1 otherwise, for a fixed quantile parameter tau in [0, 1]. It has been shown that Support Vector Machines (SVMs) in the limit are quantile classifiers with tau = 0.5. In this paper we show that by using asymmetric cost of misclassification SVMs can be appropriately extended to recover, in the limit, the quantile binary classifier for any tau. We then present a principled algorithm to solve the extended SVM classifier for all values of tau simultaneously. This has two implications: First, one can recover the entire conditional distribution P(Y=1|X=x) = tau for tau in [0, 1]. Second, we can build a risk-agnostic SVM classifier where the cost of misclassification need not be known apriori. Preliminary numerical experiments show the effectiveness of the proposed algorithm.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
Theory & Algorithms
ID Code:5484
Deposited By:Jin Yu
Deposited On:29 October 2009