Keywords

One-Class Classification, LS-SVDD, Deep Learning

Subject Categories

Applied Mathematics | Artificial Intelligence and Robotics | Computer Sciences

Abstract

One-Class Classification (OCC) focuses on learning the characteristics of normal data and identifying observations that deviate from this learned pattern as anomalies. It is commonly used in applications such as medical diagnosis, cybersecurity, industrial monitoring, and fraud detection, where abnormal examples are often rare or unavailable during training. Classical approaches such as SVDD and LS-SVDD describe normal data using a hypersphere. While effective in some settings, these methods rely on shallow representations and can be sensitive to noise and contaminated observations. To address these limitations, this dissertation introduces a Deep LS-SVDD framework that combines hypersphere-based data description with deep neural networks. By learning informative feature representations, the model can better capture complex nonlinear patterns in the data.To improve robustness, a correntropy-induced loss function is incorporated into the Deep LS-SVDD framework. Because correntropy is bounded, it naturally reduces the influence of outliers and contaminated observations during training. A half-quadratic reformulation is employed to facilitate optimization, leading to an alternating procedure that iteratively updates the auxiliary variables, hypersphere parameters, and neural network parameters until convergence. The proposed robust framework is further extended to the classical SVDD formulation, preserving the geometric interpretation of hypersphere-based methods while enhancing robustness to contamination. Simulation studies and real-world applications demonstrate that the proposed methods achieve lower Type II error rates and improved anomaly detection performance, particularly in noisy and contaminated environments. Overall, this dissertation establishes a robust and scalable deep learning framework for one-class classification that combines representation learning with principled robust optimization.

Completion Date

2026

Semester

Summer

Committee Chair

Edgard Maboudou

Degree

Doctor of Philosophy (Ph.D.)

College

College of Sciences

Department

School of Data, Mathematical, and Statistical Sciences

Format

PDF

Document Type

Dissertation

Language

English

Download

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.