PhD Oral Exam - Ali Hamid Muthanna Al-Gumaei, Information and Systems Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

As the amount of data being generated keeps growing, there is an increasing demand for adaptable approaches that can effectively extract the overall trends while also maintaining subject-specific information from large-scale datasets. Modeling complex and high-dimensional data presents significant challenges across healthcare, human action recognition, and energy disaggregation. In healthcare, analyzing functional magnetic resonance imaging (fMRI) data to uncover patterns linked to neurological disorders like Alzheimer’s and schizophrenia requires methods that account for subject variability and generalize findings across populations, enabling improved diagnostics and therapies. Similarly, human action recognition must address challenges such as movement diversity, environmental factors, and sensor noise to enhance applications like safety monitoring, assistive technologies, and personalized healthcare. In energy disaggregation, breaking down aggregate consumption data into appliance-level usage demands robust models capable of handling overlapping patterns, limited labeled data, and computational constraints, empowering energy optimization and efficiency. Across these domains, innovative data-driven methods are critical for extracting meaningful insights, driving advancements in diagnostics, behavioral analysis, and sustainable energy management.

In this dissertation, we address these challenges in high-dimensional analysis through advancements in source separation techniques. First, we develop a bounded multivariate generalized Gaussian mixture model (BMGGMM) integrated with independent component analysis (ICA) to effectively capture the correlated features in multivariate data. While independent vector analysis (IVA) extends ICA to handle multiple datasets by leveraging inter-dataset dependencies and preserving their correlation structures, it suffers from limitations when dealing with complex datasets. To overcome this, we propose a novel blind source separation (BSS) method that combines IVA with the BMGGMM framework, enabling robust modeling of complex data distributions with varying shapes and dimensions. Second, we introduced the integration of the ICA-BMGGMM and IVABMGGMM to the hidden Markov model (HMM) to boost their performance in terms of source separation for the human action recognition. The performance of IVA deteriorates as the number of datasets and sources increases and as the level of correlation across datasets decreases, particularly when the number of samples is fixed. Consequently, applying IVA to large-scale fMRI data often yields suboptimal results. To address this limitation, we propose the adaptive constrained IVA (acIVAMGGMM) and bounded acIVAMGGMM techniques. These methods integrate multiple reference signals into the IVA cost function and employ an adaptive parameter-tuning mechanism to dynamically balance the influence of accurate and inaccurate references. Through extensive evaluation on high-dimensional datasets, we demonstrate that these methods achieve superior performance, effectively extracting meaningful patterns from large-scale fMRI datasets. Finally, we introduce a new approach, ICA and IVA for common subspace analysis (ICABMGGMM-CS) and IVABMGGMM-CS, designed for the subspace analysis of multi-subject fMRI data. These methods leverage the strengths of both ICA and IVA while effectively addressing the challenges posed by high dimensions. Our results demonstrate that IVABMGGMM-CS successfully identifies meaningful common subspaces. These features enable the detection of significant subgroups among individuals with mental health disorders, offering valuable insights into their underlying neural patterns.