ORCID Identifier(s)

0009-0008-1842-5345

Graduation Semester and Year

Fall 2024

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Department

Electrical Engineering

First Advisor

Dr. Ioannis Schizas

Second Advisor

Dr. Michael Manry

Third Advisor

Dr. Qilian Liang

Fourth Advisor

Dr. Jonathan Bredow

Abstract

The increasing demand for real-time analysis of sensor data in dynamic environments necessitates innovative approaches to data clustering. This work introduces a novel Online Kernel Clustering (OKC) framework that efficiently determines time-varying clustering configurations without requiring training data. The proposed method employs sparse kernel factorization, guided by a time-dependent metric to quantify the closeness of kernel similarity matrices to a block diagonal structure. By processing data sequentially, the OKC framework is tailored for non-stationary settings. The optimization process integrates block coordinate descent, difference-of-convex functions minimization, and projected sub-gradient descent to iteratively update kernel covariance matrices and cluster memberships online. Numerical experiments using synthetic and real-world datasets confirm the superior clustering accuracy and reduced runtime of the OKC framework compared to existing methods. Additionally, a function approximating Q-learning approach enables an agent to iteratively refine kernel parameters, leveraging a state-action framework to optimize clustering outcomes dynamically. The integration of neural networks for Q-function approximation facilitates scalability and robustness in high-dimensional state-action spaces. The reinforcement learning component demonstrates adaptability to dynamic configurations, reducing computational overhead while maintaining high performance. This research contributes a robust, scalable solution for unsupervised clustering in evolving environments, offering practical applications in fields such as activity recognition and hyperspectral imaging.

Keywords

Kernel Clustering, Reinforcement Learning, Function Approximation, Neural Networks

Disciplines

Signal Processing

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.