PhD Final Oral Defense
Exploring Latent Structure in Data: Algorithms and Implementations by
November 14, 2014
Dr. Andreas Spanias (chair)
Dr. Trevor Thornton
Dr. Michael Goryll
Dr. Konstantinos Tsakalis
Feature representation for raw data is one of the most important ingredients in a machine learning system. Hand designing features is an expensive and time consuming process.
Furthermore, they do not generalize well to unseen data and novel tasks. This dissertation focuses on building data-driven unsupervised models for analyzing raw data and developing efficient feature representations. Simultaneous segmentation and feature extraction approaches for silicon-pores sensor data are considered. Algorithms to improve transform domain features for ion-channel time-series signals based on matrix completion are presented. The improved features achieve better performance in classification tasks and in reducing the false alarm rates when applied to analyte detection. We will also present a new framework for feature extraction for challenging natural environment sounds and images. Several algorithms are proposed that perform supervised tasks such as recognition and tag annotation. Several strategies to speed-up
Orthogonal Matching Pursuit algorithm using CUDA kernel on a GPU are proposed.