Online Program Home
My Program

Abstract Details

Activity Number: 594 - Methods for Analysis of High-Dimensional Data
Type: Contributed
Date/Time: Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #328780 Presentation
Title: Kernel-Based Nonlinear Dimension Reduction for Automatic Gender Classification
Author(s): Katherine Kempfert* and Yishi Wang and Cuixian Chen
Companies: University of Florida and University of North Carolina Wilmington and University of North Carolina Wilmington
Keywords: gender classification; dimension reduction; biometrics; nonlinear; kernel; image

Image data is often nonlinear and high-dimensional, which can hinder statistical and machine learning algorithms. Therefore, in this study the following nonlinear kernel-based dimensionality reduction techniques are investigated: kernel principal component analysis (KPCA), supervised kernel principal component analysis (SKPCA), and kernel Fisher's discriminant analysis (KFDA). As a preliminary, these techniques are studied on three simulated datasets. Then a novel machine learning pipeline is proposed for the longitudinal face aging database MORPH-II. First, images in MORPH-II are preprocessed, and several feature types (local binary patterns, histogram of oriented gradients, and biologically-inspired features) are extracted from the preprocessed images. The KPCA, SKPCA, and KFDA techniques are used to transform then reduce the dimension of the extracted features. The transformed, reduced dimension data serve as input for a linear support vector machine (SVM) to classify gender of the pictured subjects. Finally, the dimension reduction techniques are compared in terms of gender classification results, which achieve up to 94% testing accuracy.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program