Multi-Task Learning for Personalized Age Estimation

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

Title: "Multi-Task Learning for Personalized Age Estimation"

by

Mr. Gao Xiang

Abstract

Automatic age estimation is the problem of using a computer to predict the 
age of a person automatically based on a given facial image. This problem 
has numerous real-world applications. However, the problem is very 
challenging due both to the high variability of the aging functions of 
different people and to the sparsity of data available for model training. 
In this thesis, instead of learning a global aging function, we learn 
multiple aging functions for different people and take a multi-task 
learning approach to deal with the data sparsity issue. Our model is a 
multi-task extension of the support vector regression model. To deal with 
the sparsity of training data, we propose a similarity measure for 
clustering the aging functions. During the testing stage which involves a 
new person with no data used for model training, we propose a 
feature-based similarity measure for characterizing the test case. We have 
conducted some simulation experiments on the FG-NET and MORPH databases to 
compare our method with some state-of-the-art methods.


Date            :       12 May 2011 (Thursday)

Time            :       3:00pm to 3:40pm

Venue		: 	Room 3301A (17-18 lift)

Advisor         :	Professor D.Y. Yeung

2nd reader      :	Dr. James Kwok