The Catholic University of America

Seminar in Functional Analysis
and Related Areas

Wednesday, November 28, 2018
4:45 - 6:30 p.m. (including a coffee-break)
Aquinas Hall, Room 108

Speaker: Professor Vera Kurkova, Institute of Computer Science, Czech Academy of Sciences, Prague, Czech Republic

Title: Probabilistic Bounds for Binary Classification of Large Data Sets

Abstract: We investigate a probabilistic model for classification task relevance, deriving correlations between randomly chosen functions and network input-output functions. Impact of large data sets is analyzed from the point of view of the concentration of measure phenomenon. The Azuma-Hoeffding inequality is used, which allows applications to cases which do not satisfy the naive Bayes assumption (i.e., cases when assignments of class labels to feature vectors are not independent).T

V. Bogdan, The Catholic University of America
A. Levin, The Catholic University of America
R. Kalpathy, The Catholic University of America
P. Kainen, Georgetown University
A. Vogt, Georgetown University

Telephone: 202-319-5221
Fax: 202-319-5231