Evaluating dimensionality reduction techniques for visual category recognition using rényi entropy 19th European Signal Processing Conference (EUSIPCO 2011)

Author(s): Ashish Gupta ; Richard Bowden
Publisher: IEEE - Institute of Electrical and Electronics Engineers, Inc.
Publication Date: 1 August 2011
Conference Location: Barcelona, Spain
Conference Date: 29 August 2011
Page(s): 913 - 917
ISSN (Paper): 2076-1465



Visual category recognition is a difficult task of significant interest to the machine learning and vision community. One of the principal hurdles is the high dimensional feature space. This paper evaluates several linear and non-linear dimensionality reduction techniques. A novel evaluation metric, the rényi entropy of the inter-vector euclidean distance distribution, is introduced. This information theoretic measure judges the techniques on their preservation of structure in lower-dimensional sub-space. The popular dataset, Caltech-101 is utilized in the experiments. The results indicate that the techniques which preserve local neighborhood structure performed best amongst the techniques evaluated in this paper.