Skip to the content.

Data-Driven Model Reduction, Scientific Frontiers, and Applications ()

Shahin Shahrampour, Department of Industrial & Systems Engineering
Data-dependent Kernel Approximation for Better Generalization


Random features provide a practical framework for large-scale kernel approximation and supervised learning. In this talk, we discuss how data-dependent sampling of features can significantly reduce the prediction error in supervised learning. We present some of our recent results in this direction, including algorithmic, theoretical, and empirical findings in the past two years.