|
Activity Number:
|
436
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
| Abstract - #305087 |
|
Title:
|
Supervised Dimensionality Reduction on the Fusion of Dissimilarity Matrices
|
|
Author(s):
|
Zhiliang Ma*+ and Carey E. Priebe
|
|
Companies:
|
Johns Hopkins University and Johns Hopkins University
|
|
Address:
|
320A Clark Hall, Baltimore, MD, 21218,
|
|
Keywords:
|
dimensionality reduction ; dissimilarity ; embedding ; fusion
|
|
Abstract:
|
We explore three possible ways of combining multiple dissimilarity representations in order to obtain superior performance in statistical pattern recognition, and investigate in detail one of them---combination in the space of the Cartesian product of the embeddings. The high dimensionality of the Cartesian product space is the main obstacle to this approach. For concreteness, we consider the inferential task at hand to be classification. We propose a supervised dimensionality reduction method, which utilizes the class label information, to help achieve a favorable combination. The simulation and real data results show that our fusion approach can improve classification performance compared to the alternatives of principal components analysis and no dimensionality reduction at all.
|
