Talk by Daniel S Yeung on “Sensitivity Based Generalization Error for Single and Multiple Classifier Systems with Applications”
Sensitivity Based Generalization Error for Single and Multiple Classifier Systems with Applications
Abstract
Generalization error model provides a theoretical support for a classifier’s performance in terms of prediction accuracy. However, existing models give very loose error bounds. This explains why classification systems generally rely on experimental validation for their claims on prediction accuracy. In this talk we will revisit this problem and explore the idea of sensitivity measure in developing a new generalization error model based on the assumption that only prediction accuracy on unseen points in a neighborhood of a training point will be considered, since it will be unreasonable to require a classifier to accurately predict unseen points “far away” from training samples. Relationship between the new model and the regularization technique will be examined and a number of generic as well as domain specific applications will be presented.
Daniel S Yeung, Chair Professor, School of Computer Science and Engineering, South China University of Technology, Guangzhou, China, Junior Past President, IEEE Systems, Man and Cybernetics Society, Fellow of IEEE