PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Adaptive and non-adaptive perceptron-like algorithms for selective sampling tasks
Claudio Gentile
Foundations of Active Learning workshop at NIPS 2005 2005.

Abstract

A selective sampling algorithm is a learning algorithm for classification that, based on the past observed data, decides whether to sample the label of each new instance to be classified. In this paper we compare both theoretically and empirically two Perceptron-like selective sampling algorithms for binary classification. The empirical comparison is carried out on two well-known medium-size real-world datasets for OCR tasks. The outcome of these experiments (essentially predicted by the theoretical analysis) shows that our selective sampling algorithms tend to perform as good as the algorithms receiving the true label after each classification, while observing in practice substantially fewer labels.

Other (gzipped postscript)
EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Computational, Information-Theoretic Learning with Statistics
Learning/Statistics & Optimisation
ID Code:1288
Deposited By:Claudio Gentile
Deposited On:28 November 2005