Alternative approximation method for learning multiple feature

Abstract

© 2016 by the Mathematical Association of Thailand. All rights reserved. The theory of reproducing kernel Hilbert space (RKHS) has recently appeared as a powerful framework for the learning problem. The principal goal of the learning problem is to determine a functional which best describes given data. Recently, we have extended the hypercircle inequality to data error in two ways: First, we have extended it to circumstance for which all data is known within error. Second, we have extended it to partially-corrupted data. That is, data set contains both accurate and inaccurate data. In this paper, we report on further computational experiments by using the material from both previous work.