How to optimize sample in active learning : Dispersion, an optimum criterion for classification ?
Benoit Gandar, Gaelle Loosli and Guillaume Deffuant
Europeen conference ENBIS 2009.

Abstract

We want generate learning data appropriated to classification problems. First, we show that theorical results about low discrepancy sequences in regression problems are not adequate for classification problems. Then, we show with theorical and experimental arguments that minimising the dispersion of the sample is a relevant strategy to optimize performance of classification learning.

[Edit]