Using the Nonsmooth Analysis in a Learning Process

Autoři: Marcel Jiřina
Rok vydání: 2010
Typ publikace: Konferenční příspěvek (mezinárodní konf.)
Periodikum: Proceeding of The 2010 IRAST International Congress on Computer Applications and Computational Science
Projekty: Centrum aplikované kybernetiky
Bibliografická citace: JIŘINA, Marcel, JIŘINA JR., M. Using the Nonsmooth Analysis in a Learning Process. In: CHELLAPPAN, S., CHENG, A.C., MIN, M., QUANG, V.N, RAMALINGAM, P., YANG, H., eds. Proceeding of The 2010 IRAST International Congress on Computer Applications and Computational Science. -: IRAST, 2010, S. 91-94. ISBN 978-981-08-6846-8.
ID: 355020

Abstrakt

We propose an unconventional updating algorithm for weighting features for a classification task with abrupt changes of the error function. Using a so-called nonsmooth analysis we prove the quadratic convergence of such a learning process. For the dynamic optimization of the step size we use approach similar to Runge’s method of a half step. Finally we demonstrate its classification abilities on artificial as well as on real-life classification tasks.