DocumentCode :
75988
Title :
Information-Theoretic Dictionary Learning for Image Classification
Author :
Qiang Qiu ; Patel, Vishal M. ; Chellappa, Rama
Author_Institution :
Dept. of Electr. & Comput. Eng., Duke Univ., Durham, NC, USA
Volume :
36
Issue :
11
fYear :
2014
fDate :
Nov. 1 2014
Firstpage :
2173
Lastpage :
2184
Abstract :
We present a two-stage approach for learning dictionaries for object classification tasks based on the principle of information maximization. The proposed method seeks a dictionary that is compact, discriminative, and generative. In the first stage, dictionary atoms are selected from an initial dictionary by maximizing the mutual information measure on dictionary compactness, discrimination and reconstruction. In the second stage, the selected dictionary atoms are updated for improved reconstructive and discriminative power using a simple gradient ascent algorithm on mutual information. Experiments using real data sets demonstrate the effectiveness of our approach for image classification tasks.
Keywords :
entropy; gradient methods; image classification; image reconstruction; learning (artificial intelligence); optimisation; dictionary atom selection; dictionary compactness; dictionary discrimination; dictionary reconstruction; discriminative power improvement; entropy; gradient ascent algorithm; image classification tasks; information maximization principle; information-theoretic dictionary learning; mutual information measure maximization; object classification tasks; reconstructive power improvement; Atomic measurements; Dictionaries; Entropy; Government; Image reconstruction; Kernel; Mutual information; Dictionary learning; entropy; image classification; information theory; mutual information;
fLanguage :
English
Journal_Title :
Pattern Analysis and Machine Intelligence, IEEE Transactions on
Publisher :
ieee
ISSN :
0162-8828
Type :
jour
DOI :
10.1109/TPAMI.2014.2316824
Filename :
6787085
Link To Document :
بازگشت