Abstract

The present doctoral dissertation focuses on representative-based processing proper for a big set of high-dimensional data. Compression and subset selection are considered as two main effective methods for representing a big set of data by a much smaller set of variables. Compressive sensing, matrix singular value decomposition, and tensor decomposition are employed as powerful mathematical tools to analyze the original data in terms of their representatives. Spectrum sensing is an important application of the developed theoretical analysis. In a cognitive radio network (CRN), primary users (PUs) coexist with secondary users (SUs). However, the secondary network aims to characterize PUs in order to establish a communication link without any interference with the primary network. A dynamic and efficient spectrum sensing framework is studied based on advanced algebraic tools. In a CRN, collecting information from all SUs is energy inefficient and computationally complex. A novel sensor selection algorithm based on the compressed sensing theory is devised which is compatible with the algebraic nature of the spectrum sensing problem. Moreover, some state-of-the-art applications in machine learning are investigated. One of the main contributions of the present dissertation is the introduction a versatile data selection algorithm which is referred as spectrum pursuit (SP). The goal of SP is to reduce a big set of data to a small-size subset such that the linear span of the selected data is as close as possible to all data. SP enjoys a low-complexity procedure which enables SP to be extended to more complex selection models. The kernel spectrum pursuit (KSP) facilitates selection from a union of non-linear manifolds. This dissertation investigates a number of important applications in machine learning including fast training of generative adversarial networks (GANs), graph-based label propagation, few shot classification, and fast subspace clustering.

Notes

If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu.

Graduation Date

2021

Semester

Summer

Advisor

Rahnavard, Nazanin

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Electrical and Computer Engineering

Degree Program

Electrical Engineering

Format

application/pdf

Identifier

CFE0008674;DP0025405

URL

https://purls.library.ucf.edu/go/DP0025405

Language

English

Release Date

August 2021

Length of Campus-only Access

None

Access Status

Doctoral Dissertation (Open Access)

Share

COinS