Feature Reduction Based on Sum-of-SNR (SOSNR) Optimization
Paper in proceedings, 2014

Dimensionality reduction plays an important role in machine learning techniques. In classification, data transformation aims to reduce the number of feature dimensions, whereas attempts to enhance the class separability. To this end, we propose a new classifier-independent criterion called 'Sum-of-Signal-to-Noise-Ratio' (SoSNR). A framework designed for maximization with respect to this criterion is presented and three types of algorithms, respectively based on (1) gradient, (2) deflation and (3) sparsity, are proposed. The techniques are conducted on standard UCI databases and compared to other related methods. Results show trade-offs between computational complexity and classification accuracy among different approaches.

classification

Fisher's Score

Sum-of-SNR

SODA

feature reduction

Author

Yinan Yu

Chalmers, Signals and Systems, Signalbehandling och medicinsk teknik, Signal Processing

Tomas McKelvey

Chalmers, Signals and Systems, Signalbehandling och medicinsk teknik, Signal Processing

S.Y. Kung

Princeton University

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

15206149 (ISSN)

6756-6760 6854908

Areas of Advance

Information and Communication Technology

Subject Categories

Electrical Engineering, Electronic Engineering, Information Engineering

DOI

10.1109/ICASSP.2014.6854908

ISBN

978-147992892-7

More information

Created

10/7/2017