Computer Sciences Commons^™

From Neural Networks To Large Language Models: Innovations In Financial Ai, Mathematical Reasoning, And Structured Data Representation, Junyi Ye

Dissertations

This dissertation explores the evolution and application of artificial intelligence techniques across three critical domains: financial modeling, mathematical reasoning, and structured data analysis. The dissertation presents seven research projects that chart a progression from specialized neural architectures to sophisticated large language models (LLMs), contributing novel methodologies and frameworks at each stage.

In the financial domain, the research first introduces TS-Mixer, a MLP-based architecture for time-series forecasting that captures both feature relationships and temporal dependencies through a simple yet effective design, outperforming more complex models in S&P500 index prediction. The dissertation then presents DySTAGE, a dynamic graph representation learning framework that …

Towards Explainable Ai On Graph Neural Networks: Xaig, Jiaxing Zhang

Dissertations

In the evolving landscape of artificial intelligence (AI), Graph Neural Networks (GNNs) have garnered growing prominence for their adeptness in processing graph-structured data. Despite this, the interpretability of their predictions often remains elusive. The demand for transparency and explainability in complex prediction models has reached unprecedented levels. To address this, post-hoc instance-level explanation techniques have emerged, aiming to unveil the rationale behind GNN predictions. These techniques endeavor to unearth substructures that elucidate the predictive behavior of trained GNNs.

This dissertation embarks on an exploration of Explainable AI (XAI) technologies within the realm of GNNs. Amid the challenges posed by the …

Abel Inversion Comparison Of Geant4 Simulation And Ozone Production Using Cavity Ringdown Spectroscopy In Nitrogen/Oxygen Mixtures In The Presence Of Alpha Radiation, Sidney John Gautrau

Dissertations

The effects of radioactive materials on atmospheric gases have been a topic of interest for years. Radioactive materials ionize the surrounding air, and subsequent reactions lead to molecules such as ozone and nitrogen oxides. The presence of these species above background levels can be used as a marker for radioactive materials which has desirable defense applications like remote detection of radioactive materials. The molecules created in the presence of radioactive materials have been quantified in literature using G-values, which is the number of molecules of a product produced per 100 eV of deposited energy. In this work, Cavity Ringdown Spectroscopy …

Resilient Learning For Anomaly Detection In Smart Living Systems, Sahar Abedzadeh

Dissertations

Cyber-Physical Systems (CPS) rely on anomaly-based detection methods to ensure the integrity and security of critical infrastructures such as smart grids, smart water metering systems, and advanced metering infrastructures (AMI). Anomaly detection methods are commonly used to identify deviations from normal system behavior by establishing learned profiles and thresholdbased distinctions between benign and anomalous events. However, conventional frameworks often fail to account for adversarial data poisoning attacks, unlabeled unsafe events, and environmental noise—factors that distort training data, degrade detection accuracy, and increase false alarms. This dissertation proposes a resilient learning framework that mitigates these biases by integrating quantile regression, M-estimation …

Overcoming Motor Imagery Bci Illiteracy: Adaptive Decoding And Knowledge Transfer In Eeg-Based Brain-Computer Interfaces, Zaid Shuqfa

Dissertations

Brain–computer interfaces (BCIs), also known as brain–machine interfaces (BMIs), enable direct communication between the brain and external devices without the involvement of peripheral nerves or muscles. Among various BCI paradigms, motor imagery (MI)–based BCIs are particularly appealing due to their intuitive, cue-independent nature, allowing users to issue control commands at will. MI–BCIs hold substantial promise for improving the quality of life of individuals with motor impairments, as well as enhancing hands-free control for healthy users. However, their widespread adoption remains limited by challenges such as low signal-to-noise ratio, inter- and intra-subject variability, and the need for frequent calibration. These challenges …

Pushing The Boundaries Of Large Language Models: Innovations And Limitations In Nlp, Finance, And Mathematics, A M Muntasir Rahman

Dissertations

Large Language Models (LLMs) have emerged as transformative tools across a spectrum of domains, yet their practical deployment reveals a blend of remarkable potential and notable limitations. This research explores innovative methodologies to extend the capabilities of LLMs while addressing critical challenges in their evaluation and application. By leveraging rule-based approaches, the in-context learning capabilities of LLMs, and human-in-the-loop validation across three focused studies, this research introduces robust strategies for dataset synthesis, model enhancement, and model assessment in three distinct domains: natural language processing, financial sentiment analysis, and mathematical reasoning

The first study proposes an efficient data augmentation framework, EASE, …

First-Principles Study Of Ferroelectric Properties And Co2 Reduction Reaction Capabilities In Two-Dimensional Monolayers And Heterostructures, Mo Li

Dissertations

Two-dimensional (2D) materials hold significant potential for CO₂ reduction reactions (CO₂RR) due to their high surface-to-volume ratio. However, achieving high selectivity for desired products and overcoming limitations posed by scaling relationships remain challenging. Recent studies suggest that ferroelectric (FE) materials with switchable out-of-plane polarization (OOP) can effectively tune the adsorption behavior, thermodynamics, and kinetics of CO₂RR, offering promising solutions to these challenges. Using density functional theory (DFT) and the Berry phase approach, this work expands the family of 2D ferroelectrics by theoretically identifying Y₂CO₂, Y₂CS₂, and Sc …

Ensemble Learning Models For Large-Scale Time Series Forecasting In Supply Chain, Minjuan Zhang

Dissertations

Machine learning and AI techniques are transforming supply chain forecasting, driven by the expanding availability of data assets. These advanced methods offer powerful opportunities to optimize management processes, reduce operational costs, and enhance strategic decision-making, which is crucial for enterprise success. However, conventional statistical approaches, such as Autoregressive Integrated Moving Average Models (ARIMA), dynamic regression, and Unobserved Component Models (UCMs)—which have long dominated time series forecasting—often fall short in accuracy and scalability. These traditional models face limitations in batch processing, handling large-scale data, addressing uncertainty-induced disruptions, and synchronizing demand-supply scenarios.

To address these challenges, a novel class of AI-powered ensemble …

Machine Learning Methods For Pattern Recognition Analysis Of Genomic And Molecular Data, Kuang Du

Dissertations

While immune therapies achieve remarkable success in treating various cancers, only a subset of patients achieves a durable clinical response, and many exhibit innate or acquired resistance. Precision medicine aims to tailor treatments to individual patients based on specific biological markers, ensuring that each patient receives the therapy most likely to be effective. Predictive biomarkers and gene signatures offer potential for more personalized treatment strategies by identifying patients likely to benefit. Recent studies suggest that gene signatures, comprising sets of genes, hold predictive value for certain clinical variables. Typically derived from biological expert knowledge, these signatures demonstrate substantial predictive potential, …

Knowledge Diffusion In Networks Of Artificial Learners, Ehsan Beikihassan

Dissertations

The dissertation draws inspiration from the topic of peer learning in the social sciences and the study of information dissemination and knowledge diffusion in network science. In particular, it introduces and studies a setting involving a population or network of artificial learners, with the objective of optimizing aggregate performance measures under constraints on training resources. In this context, natural knowledge diffusion processes in networks of interacting artificial learners are studied. The term "natural" refers to processes that emulate human peer learning, where the internal state and learning processes of students remain largely opaque, and the main degree of freedom lies …

Interactive Visualization Workflows For Mitigating Analytical Uncertainty, Kaustav Bhattacharjee

Dissertations

This dissertation takes a process-centric and stakeholder-first perspective for handling analytical uncertainty: the form of uncertainty that confronts data analysts' insight-generation processes in high-consequence decision-making scenarios. The cost of an incorrect decision when data is used for movie recommendations as opposed to when personal data is used to drive insights or when data-driven modeling is used to drive real-time decisions for maintaining the health of a grid are vastly different in terms of consequences. This dissertation looks at analytical uncertainty in two real-world scenarios: i) how sensitive information leakage can be prevented during the open data release process with data …

Crowd-Sourced Learning For Computer Graphics Applications, Yunhao Zhang

Dissertations

Computer Graphics (CG) revolves around virtual content creation using computational methods, spanning applications from games to visual effects. Typically, the creation of CG content is led by expert practitioners who guide computational algorithms towards satisfactory results. Thus, creating CG content often requires manual iterations encompassing algorithm design, parameter tuning, and aesthetic feedback. This work investigates how to leverage crowd-sourcing to streamline such creation processes, focusing on animation and simulation. In animation, a novel crowd-sourcing framework is proposed for combat animation, enabling users to analyze motion similarities, and retrieve matching motions using novel crowd-sourced motion features. Such features enable quantifying previously …

Tackling Toxicity And Harassment In Online Environments Through The Use Of Artificial Intelligence, Heba Saleous

Dissertations

With the increase in popularity of online communities, such as social media platforms, online games, and chatroom servers, there is a need to improve chat and content moderation. Platforms have reported an increase in the prevalence of toxic behavior and hate speech. Meanwhile, moderators are reporting difficulties in keeping up with the amount of data to check as well and the type of content they are exposed to, which further harms their own mental health. The main objective of this work is to address the challenges that exist within online communities with the rising prevalence of hate speech. Additionally, some …

Multi-Label Classification Using Conformal Prediction, Chhavi Tyagi

Dissertations

In many machine learning applications, such as image tagging, document classi-fication, and medical diagnosis, a data instance can be associated with multiple classes in parallel so that each instance is associated with multiple response variables simultaneously defining multi-label classification. Standard multi-label classification methods that provide point predictions have been developed. They lack in quantifying the uncertainty of predictions. These methods also lack in accounting for label dependencies and are very computationally expensive. This dissertation develops two methods of multi-label classification using conformal prediction that quantify the uncertainty of predictions. Chapter 1 introduces notations and tools that have been used in …

Certifying Stability In Runge-Kutta Schemes: Algebraic Conditions And Semidefinite Programming, Austin Juhl

Dissertations

Numerical stability is a critical property for a time-integration scheme. In the context of Runge-Kutta methods applied to stiff differential equations, A-stability is one of the most basic and practically important notions of stability. Dating back to the work of Dahlquist, it has been known that A-stability is equivalent to the Runge-Kutta stability function satisfying a particular convex feasibility problem. Specifically, up to a transformation, the stability function lies in the convex cone of positive functions. In recent years, sum-of-squares optimization and semidefinite programming have become valuable tools in developing rigorous certificates of stability in dynamical systems. Therefore, it is …

Federated Learning Systems For Mobile Sensing Data, Xiaopeng Jiang

Dissertations

Federated Learning (FL) has emerged as a new distributed Deep Learning (DL) paradigm that enables privacy-aware training and inference on mobile devices with help from the cloud. This dissertation presents a comprehensive exploration of FL with mobile sensing data, covering systems, applications, and optimizations.

First, a mobile-cloud FL system, FLSys, is designed to balance model performance with resource consumption, tolerate communication failures, and achieve scalability. In FLSys, different DL models with different FL aggregation methods can be trained and accessed concurrently by different apps. In addition, FLSys provides advanced privacy-preserving mechanisms and a common API for third-party app developers to …

A Methodological Framework For Ontology Development, Enrichment, And Application In Natural Language Processing Tasks, Navya Martin Kollapally

Dissertations

Electronic Health Records (EHRs) have been widely used in healthcare to record demographics, vital signs, test results, immunizations, medical imaging reports, differential diagnoses, etc. It is now accepted that non-clinical (e.g., social) factors have a substantial influence on health outcomes. Hence, it is desirable to record these Social and Commercial Determinants of Health (SDoH & CDoH) in an EHR. The "non-text parts" of EHR notes (e.g., data tables) rely on coded terms from underlying ontologies or terminologies to facilitate semantic interoperability. Ontologies help define concepts, the relationships between them, and instances that can be utilized in research.

The first accomplishment …

Numerical Techniques For Improving Simulations Of Tropical Cyclones, Yassine Tissaoui

Dissertations

The increasing frequency and intensity of tropical cyclones (TCs) due to climate change pose significant challenges for forecasting and mitigating their impacts. Despite advancements, accurately predicting TC rapid intensification (RI) remains a challenge. Large eddy simulation (LES) allows for explicitly resolving the large eddies involved in TC turbulence, thus providing an avenue for studying the mechanisms behind their intensification and RI. LES of a full tropical cyclone is very computationally expensive and its accuracy will depend on both explicit and implicit dissipation within an atmospheric model. This dissertation presents two novel numerical methodologies with the potential to improve the efficiency …

An Experimental Study Of Supervised Machine Learning Techniques For Minor Class Prediction Utilizing Kernel Density Estimation: Factors Impacting Model Performance, Abdullah Mana Alfarwan

Dissertations

This dissertation examined classification outcome differences among four popular individual supervised machine learning (ISML) models (logistic regression, decision tree, support vector machine, and multilayer perceptron) when predicting minor class membership within imbalanced datasets. The study context and the theoretical population sampled focus on one aspect of the larger problem of student retention and dropout prediction in higher education (HE): identification.

This study differs from current literature by implementing an experimental design approach with simulated student data that closely mirrors HE situational and student data. Specifically, this study tested the predictive ability of the four ISML classification models (CLS) under experimentally …

Empirical Exploration Of Software Testing, Samia Alblwi

Dissertations

Despite several advances in software engineering research and development, the quality of software products remains a considerable challenge. For all its theoretical limitations, software testing remains the main method used in practice to control, enhance, and certify software quality. This doctoral work comprises several empirical studies aimed at analyzing and assessing common software testing approaches, methods, and assumptions. In particular, the concept of mutant subsumption is generalized by taking into account the possibility for a base program and its mutants to diverge for some inputs, demonstrating the impact of this generalization on how subsumption is defined. The problem of mutant …

Network Slicing And Noma Enabled Mobile Edge Computing For Next-Generation Networks, Mohammad Arif Hossain

Dissertations

The advent of next-generation wireless networks ushers in a new era of potential, harnessing cutting-edge technologies like mobile edge computing (MEC), non-orthogonal multiple access (NOMA), and network slicing as pivotal drivers of transformation. Within this landscape, an innovative approach is proposed by introducing a NOMA-enabled network slicing technique within MEC networks. This approach aims to achieve multiple objectives: meeting stringent quality of service requirements, minimizing service latency, and enhancing spectral efficiency. By seamlessly integrating NOMA with network slicing in edge computing environments, significant reductions in overall latency are achieved, alongside ensuring optimal resource allocation for NOMA users. To address these …

Information Theoretic Bounds For Capacity And Bayesian Risk, Ian Zieder

Dissertations

In this dissertation, the problem of finding lower error bounds on the minimum mean-squared error (MMSE) and the maximum capacity achieving distribution for a specific channel is addressed. Presented are two parts, a new lower bound on the MMSE and upper and lower bounds on the capacity achieving distribution for a Binomial noise channel. The new lower bound on the MMSE is achieved via use of the Poincare inequality. It is compared to the performance of the well known Ziv-Zakai error bound. The second part considers a binomial noise channel and is concerned with the properties of the capacity-achieving distribution. …

Sensing With Integrity: Responsible Sensor Systems In An Era Of Ai, David Eisenberg

Dissertations

Deep and machine learning now offer immense benefits for consumer choice, decision-making, medicine, mental health and education, smart cities, and intelligent transportation and driver safety. However, as communication and Internet technology further advances, these benefits have the potential to be outweighed by compromises to privacy, personal freedom, consumer trust, and discrimination. While ethical consequences for personal freedom and equity rise from these technological advances, the issue may not be the technology itself but a lack of regulation and policy that allow abuses to occur. A first study examines how emerging sensor-based technologies, limited to only accelerometer and gyroscope data from …

A Machine Learning-Assisted Steering And Scheduling Framework For Big-Data Scientific Workflows On Heterogeneous Computing Platforms, Yijie Zhang

Dissertations

In next-generation scientific applications, the exponential growth of big data necessitates advanced techniques for efficient data storage, processing, and analysis. This has led to the construction of intricate computing workflows, managed and orchestrated by powerful engines in big data systems as exemplified by Hadoop. As scientific applications increasingly shift towards simulation-centric approaches, traditional methodologies face new challenges in accommodating the complexity of extreme-scale numerical modeling with numerous tunable parameters. To address these challenges, this dissertation propose to develop a machine learning-assisted framework that enables autonomous computational steering of scientific simulations and optimized execution of big-data workflows on heterogeneous platforms. This …

Finding Combinatorial Patterns In Real Valued Omics Data, Kenneth Smith

Dissertations

Precision medicine is a healthcare approach which tailors disease prevention and treatment to an individual, based on their genetics, environment, lifestyle, and physiological state. These factors interact to produce biological changes that can be measured to produce data called omics, and include genomics, lipidomics, and proteomics. Despite the abundance of omics data and analysis techniques, researchers still struggle to identify biological findings that replicate across data sets and translate into clinical applications. In this dissertation, we employ combinatorial optimization techniques to improve upon three steps in the precision medicine analysis pipeline: 1) data cleaning, 2) community detection, and 3) feature …

A Holistic And Collaborative Behavioral Health Detection Framework Using Sensitive Police Narratives, Martin Keagan Wynne Brown

Dissertations

Identifying behavioral health is paramount for law enforcement officers to provide appropriate follow-up community care. In the current practice, law enforcement offices manually identify these behavioral health cases to allow the designation of the relevant follow-up resources. Police reports generated by officers' response to 911 calls remain an untapped resource for identifying such incidents. Therefore, we advocate for the incorporation of manual annotations from experts, natural language processing (NLP), active learning, advanced machine learning, and ensemble techniques to detect behavioral health cases within police reports. In this dissertation, we develop tools and frameworks to automatically detect behavioral health cases from …

Enhancing Rotating Machinery Fault Diagnosis: A Dual-Head Attention Mechanism In Deep Learning Neural Networks, Qing Snyder

Dissertations

Rotating machinery is crucial to production efficiency and safety in manufacturing industries for an extended time. Ensuring machinery reliability necessitates effective diagnostic systems, particularly for rotating bearings, the key components of such equipment. Fault diagnosis in rotating machinery is essential to prevent failures and minimize downtime, thereby playing an important role in industrial operations. The application of advanced neural network techniques in industry has risen recently. Among these, attention-based neural networks, especially the Transformer models, are originally noteworthy for their sequential data handling capability. This research delves into attention-based algorithms for rotating machinery fault diagnosis, signifying a substantial advancement in …

An Intelligent Framework Towards Fully Autonomous Driving Fueled By Smart Roads, Muhammad Jalal Khan

Dissertations

Autonomous Vehicles (AVs) are transforming next-generation autonomous mobility. These vehicles promise to increase road safety, improve traffic efficiency, reduce vehicle emissions, and enhance overall mobility. They achieve higher levels of Autonomous Driving (AD) by integrating sensors, communication technologies, computation, and Artificial Intelligence (AI). Apart from these advancements, significant gaps remain in integrating heterogeneous technologies and disciplines essential for optimizing AD. Therefore, the current solution approaches lack the capability to exploit intelligent road infrastructures and effectively orchestrate perceptual services for complex driving scenarios. The main objective of this dissertation is to address these challenges by proposing a novel end-to-end intelligent framework …

Enhancing Health Analytics: Secure And Private Federated Learning Solutions, Nisha Thorakkattu Madathil

Dissertations

Federated Learning (FL) is a collaborative method allowing individuals to train a model jointly without sharing their local datasets. It utilizes decentralized data sources to protect privacy, making it particularly promising in medical contexts where data confidentiality is paramount. FL facilitates the use of diverse datasets from various healthcare organizations while upholding patient confidentiality. It also plays a crucial role in advancing medical research and healthcare services while adhering to data distribution and compliance requirements. The primary challenges within federated healthcare encompass privacy preservation among sensitive distributed data, ensuring efficient communication, addressing data heterogeneity, and ultimately guaranteeing model accuracy. To …

Cybersecurity Education And Continuous Learning Towards Uae Ncsp Fulfilment, Saleh Hamad Aldaajeh

Dissertations

This dissertation delves into enhancing cybersecurity education by aligning academic curricula with national cybersecurity strategic plan (NCSP) objectives, emphasizing the crucial role of Higher Education Institutions (HEIs) in developing a skilled cybersecurity workforce. Analyzing ten NCSPs, it identifies strategic themes and gaps between national goals and HEI offerings. The study reviews NCSP guidelines, international cybersecurity indices, and literature, including the NICE-NIST framework, to develop a framework that bridges the educational gap, improving learning outcomes and arming students with vital skills, knowledge, and competencies. Furthermore, it introduces a platform for continuous cybersecurity learning, employing micro-credentials, blockchain technology, and AI-driven systems. Based …