AUTHOR=Mhaskar Hrushikesh , Pereverzyev Sergei V. , Semenov Vasyl Yu. , Semenova Evgeniya V. 

TITLE=Data Based Construction of Kernels for Semi-Supervised Learning With Less Labels

JOURNAL=Frontiers in Applied Mathematics and Statistics

VOLUME=5

YEAR=2019

URL=https://www.frontiersin.org/journals/applied-mathematics-and-statistics/articles/10.3389/fams.2019.00021

DOI=10.3389/fams.2019.00021

ISSN=2297-4687

ABSTRACT=<p>This paper deals with the problem of semi-supervised learning using a small number of training samples. Traditional kernel based methods utilize either a fixed kernel or a combination of judiciously chosen kernels from a fixed dictionary. In contrast, we construct a data-dependent kernel utilizing the Mercer components of different kernels constructed using ideas from diffusion geometry, and use a regularization technique with this kernel with adaptively chosen parameters. Our algorithm is illustrated using a few well-known data sets as well as a data set for automatic gender identification. For some of these data sets, we obtain a zero test error using only a minimal number of training samples.</p>