A0635
Title: Kernel conditional density machines
Authors: Paul Schneider - USI Lugano and SFI (Switzerland)
Damir Filipovic - EPFL and Swiss Finance Institute (Switzerland) [presenting]
Michael Multerer - USI Lugano (Switzerland)
Abstract: Applications across many different fields increasingly call for nonparametric multivariate distributions, from which in particular conditional distributions can consistently be obtained. A novel nonparametric framework for distribution estimation is proposed in reproducing kernel Hilbert spaces. Finite sample guarantees are provided consistent results along with data asymptotics. An application in conditional quantile regression shows excellent performance relative to existing approaches in particular for large sample sizes.
