Binary classification with noise condition

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Presentation

Title: "Binary classification with noise condition"

By

Xuan CHEN

Abstract

Traditionally, the study of binary classification has been formulated as a 
deterministic problem with 0-1 labels. However, probabilistic labels are 
becoming popular nowadays since they are more informative. In this final 
year thesis, the student studied the accuracy of some prediction models 
under different noise conditions when probabilistic labels were adopted. 
The conclusion shows that, with the use of probabilistic labels, the 
results in all cases were no worse than the ones when 0-1 labels were used. 
Besides, if more instances are found at the classification "boundary", the 
prediction will be less accurate.

Date:                   Monday, 27 April 2015

Time:                   6:10 - 6:50pm

Venue:                  Room 5505
                        Lifts 25/26

Committee Members:      Dr. Raymond Wong (Supervisor)
                        Prof. Dit-Yan Yeung (Reader)