Physical Sciences and Mathematics Commons^™

Counterfactual Explanations For Time Series Classification: From Localized Perturbations To Realistic Generation, Peiyu Li

All Graduate Theses and Dissertations, Fall 2023 to Present

Machine learning models are often used to classify signals collected over time, such as heart rhythms, movement recordings, industrial sensor measurements, and scientific observations. These models can be accurate, but they are often difficult to understand. Users may need to know not only what a model predicted, but also what would have needed to change for the model to reach a different decision.

This dissertation studies counterfactual explanations for time series data. A counterfactual explanation answers a “what-if” question. For example, if a model classifies a signal as one activity instead of another, the explanation shows how the signal would …

Modeling And Generating Crash Avoidance Behaviors In Safety-Critical Vehicle–Pedestrian Interactions Using Deep Reinforcement Learning, Qingwen Pu

Civil & Environmental Engineering Theses & Dissertations

Traffic crashes between vehicles and pedestrians arise from complex, split-second interactions in which both parties make rapid evasive decisions. Four fundamental gaps persist in existing research. Surrogate safety measures assume linear trajectories, failing to capture the curved movements of turning vehicles and crossing pedestrians at intersections. Single-agent modeling treats one party as a fixed obstacle, ignoring the joint decision-making that governs near-miss outcomes. The effect of vehicle type on pedestrian avoidance behavior—whether pedestrians respond differently to automated vehicles (AVs) than to human-driven vehicles (HDVs)—remains poorly understood. Finally, automated driving system development is hampered by a severe scarcity of large-scale, behaviorally …

Ai-Ready Libraries Require Ai-Ready Librarians: Building Organisational Capability For Digital Transformation, Salihin Mohammed Ali

Research Collection Library

Academic libraries worldwide are rapidly experimenting with artificial intelligence (AI) to enhance research, learning, discovery, operations, and user engagement. However, many institutions continue to approach AI adoption primarily through isolated pilots, individual experimentation, or technology-centric initiatives. While these efforts generate innovation, they often struggle to scale sustainably without corresponding organisational capability development. This presentation argues that AI-ready libraries require AI-ready librarians and proposes an organisational capability approach for sustainable AI transformation in academic libraries. Drawing from the development of a library-wide AI strategy plans at Singapore Management University, the presentation explores how AI capability-building can be operationalised across diverse functional …

Predicting Student Belonging In Computing Education: A Multimodal Machine Learning Approach Using Eeg And Survey Data, Hannah Moshtaghi

Master's Theses

Measuring students’ sense of belonging, characterized by feelings of acceptance, inclusion, and encouragement from teachers, remains a significant challenge in computing education. Prior research has associated this multidimensional construct with positive academic outcomes and has identified instructors’ growth- and fixed-mindset messaging as a potential influence. However, belonging is a complex and deeply personal experience that is difficult to capture through direct observation alone. Current measurement methods rely on self-report surveys, which may not capture every aspect of an experience that can also involve emotional and cognitive responses.

This thesis investigates whether combining EEG data recorded during a belonging questionnaire with …

The Relativity Of Education: Student Perspectives On Artificial Intelligence In Community College, Ashley Greta Magana

Electronic Theses, Projects, and Dissertations

This hermeneutic phenomenological study examined how diverse community college students experience and make meaning of the integration of generative artificial intelligence (AI) into their educational contexts. Although AI is quickly transforming higher education through automated grading, personalized learning systems, and new models of assessment, the discourse surrounding its implementation remains dominated by administrators, faculty, and institutional stakeholders, while the perspectives of students, specifically community college students who are often historically underrepresented and economically marginalized, are systematically excluded. Most existing research is quantitative and centered on universities, leaving a critical gap in qualitative understanding of the most diverse population in higher …

Theory-Informed Generative Agents For Human Behavioral Modeling In Disasters, Liming Lu

All Dissertations

This dissertation develops a theory-informed generative-agent framework for modeling human behavioral decisions in disasters. Existing flood and disaster preparedness models often emphasize physical hazards, infrastructure exposure, or statistical correlations, but they struggle to capture the heterogeneous and evolving choices households make. This limitation is especially important for climate-related hazards, where future damage depends not only on changes in rainfall, inundation, and urban development, but also on decentralized protective actions such as house elevation, flood insurance, evacuation, and early preparedness. The dissertation integrates two empirical studies: a flood-risk study in Charleston, South Carolina, and a household disaster-preparedness study across hurricane contexts. …

Leveraging Deep Learning Recurrence And Attention Mechanisms For Flood Forecasting And Assessment, Elnaz Heidari

All Dissertations

Predicting how much water will flow in rivers and streams is important for managing floods, water supply, and the environment. Traditionally, government agencies have used complex models, such as the National Water Model (NWM), which simulate how much water moves through landscapes using physical laws and real-world data. However, recent advances in Artificial Intelligence (AI) have enabled new ways to make these predictions. This research explored whether AI-based models could predict river discharge more accurately. These AI models learn patterns from past data instead of relying only on physical rules. To find out how well they work, the AI models …

Caring With Ai: The Efficacy Of A Customised Chatgpt-Delivered Self-Compassion Intervention On College Students' Well-Being And Academic Functioning, Tracy Xi Chen, Chi-Ying Cheng, Andree Hartanto

Research Collection School of Social Sciences

College students face various challenges, including academic pressure, social stress, and the transition into adulthood, which can lead to increased anxiety and other mental health issues. By recognizing personal struggles as part of a shared human experience and responding with kindness, self-compassion serves as a powerful strategy for enhancing resilience, facilitating better well-being and performance outcomes. Although effective, Compassion-Focused Therapy often requires substantial resources and time, limiting its applicability to college students. To overcome these barriers, the current study designed and evaluated Your Self-Compassion Companion, a ChatGPT-powered AI chatbot intervention grounded in self-compassion theory and delivered over three weekly 20-min …

Alfred Russel Wallace Notes 42: How Accurate Are The Transcriptions Presented At The Alfred Russel Wallace Page Website?, Charles H. Smith

Faculty/Staff Personal Papers

A look is taken at the level of accuracy displayed by the transcriptions of Wallace writings offered at the Alfred Russel Wallace Page website, as determined by a ChatGPT analysis.

A Patch-Level Framework For Urban Vegetation Water Demand Estimation Using Remote Sensing And Deep Learning, Jesus Daniel Pereyra Manriquez

Open Access Theses & Dissertations

Urban water management in semi-arid regions requires an improved understanding of how vegetation and climatic conditions influence landscape water demand. Existing approaches often lack an integrated, spatially consistent framework to quantify this relationship at fine scales. This study proposes a patch-level framework to estimate relative landscape water demand by integrating vegetation coverage, vegetation condition, and atmospheric demand. Vegetation coverage is derived from high-resolution imagery obtained from the National Agriculture Imagery Program (NAIP) using a U-Net segmentation model with a MobileNetV2 backbone. A patch-based representation is used to ensure spatial consistency across the study area. Seasonal vegetation dynamics are captured using …

Domain Adaptation Of Facial Age Estimation For Law Enforcement Mugshot Repositories, Jorge Alejandro Pacheco Roque

Open Access Theses & Dissertations

Facial age estimation supports law enforcement via image-based, age-filtered queries, age-progressive re-identification, and bulk record labeling, where prediction accuracy determines if the resulting decisions can be trusted. State-of-the-art models excel on web imagery but incur higher error on mugshots due to domain shift between the professionally lit, filtered, and posed web photographs used during pre-training and the uniform backgrounds, uncooperative expressions, and decades of evolving capture technology found in mugshot collections. We address this gap by adapting SwinFace - a state-of-the-art multi-task Swin Transformer with public code and pretrained weights, trained on color face imagery for face recognition, facial expression …

Towards A Multi-Framework Approach To Explainable Artificial Intelligence, Henry Salgado

Open Access Theses & Dissertations

Artificial Intelligence (AI) and machine learning (ML) models are increasingly being deployed to support decision-making in high-stakes domains such as healthcare, criminal justice, and education, where trust, accountability, and transparency are critical. However, increasing model complexity has made many modern systems insufficiently transparent. Existing approaches to explainable AI (XAI) typically emphasize either intrinsic model simplicity or post-hoc attribution methods that estimate feature importance for predictions. While these approaches provide valuable insights into model behavior, they do not necessarily establish whether the identified importance is grounded in the underlying data patterns or in the structural relationships that generate model behavior. Many …

Causal Discovery Methods For Single Cell Rna-Seq Data, Melanie Lambert

All Dissertations

The advancement of single cell RNA sequencing (scRNA-seq) has enabled the study of causal relationships between genes at single cell resolution. Although many causal discovery methods have been applied to scRNA-seq perturbation data, they are not well-suited to capture the characteristics of scRNA-seq data. The overall goal of this dissertation is to enhance researchers' ability to gain insight into genetic relationships.

The scRNA-seq data is high-dimensional, typically containing thousands of genes, and is sparse and zero-inflated due to dropout events, as well as noisy and subject to biological variability. Traditional causal discovery methods, such as constraint or score-based approaches, do …

Artificial Intelligence (Ai) In Forensic Psychology: An Umbrella Review Of Potentials And Pitfalls, Ysabel Thereze Ang Guevarra, Nur Eva Alisha Binte Mohamed Hisham, Andree Hartanto

Research Collection School of Social Sciences

Artificial intelligence (AI) is becoming increasingly embedded within forensic psychological practice, shaping how criminal risk, legal responsibility and public safety are assessed. AI tools are now used in recidivism prediction, behavioural analysis, deception detection and investigative support, high-stakes domains where errors can have profound consequences. Despite this rapid adoption, the existing literature remains fragmented, with most reviews confined to narrow subdomains and offering limited integrated synthesis of AI′s broader role in forensic psychology. Thus, this umbrella review addresses this gap by synthesising findings from 43 reviews obtained from five major databases, namely EBSCOhost ERIC, EBSCOhost PsycInfo, PubMed, Scopus and Web …

Exploring The Determinants Of User Discontinuance In Ai-Driven Usage-Based Insurance, Wenzhuo Li

Theses and Dissertations in Business Administration

While interest in algorithmic decision-making continues to grow, limited research has examined the post-adoption phase. This study examines how users evaluate their post-adoption experiences with algorithmic decision-making in the context of usage-based insurance (UBI), focusing on how expectation disconfirmation shapes satisfaction and the intention to discontinue use. It explores two key questions: What factors influence users’ discontinuance intention toward AI-based UBI systems? And how do specific algorithmic characteristics alter how users form these post-adoption evaluations? To investigate these questions, this study develops a comprehensive theoretical model that integrates the Expectation Confirmation Model and Reactance Theory, incorporating additional factors such as …

Automated Testing And Enhancement For Autonomous Driving Systems, Mingfei Cheng

Dissertations and Theses Collection (Open Access)

Autonomous Vehicles (AVs) rely on Autonomous Driving Systems (ADSs) to perceive their surroundings, predict the behavior of traffic participants, plan driving actions, and control vehicle motion in complex traffic environments. As AVs move toward large-scale real-world deployment, ADSs are expected not only to avoid collisions and traffic-rule violations, but also to make robust decisions, cooperate effectively with other AVs, and maintain high-quality driving behavior. However, the open, dynamic, and uncertain nature of driving environments makes the systematic testing and enhancement of ADSs a challenging problem.

Although scenario-based testing has made substantial progress in discovering safety-critical violations, several important limitations remain. …

Practical On-Device Intelligence For Resource-Constrained Embedded And Wearable Devices, Xiao Ma

Dissertations and Theses Collection (Open Access)

Embedded and wearable devices increasingly rely on deep neural networks to support intelligent sensing applications directly on the device. However, building practical on-device intelligence remains challenging throughout the deployment lifecycle. Before deployment, models must remain accurate despite severe memory, computation, and energy budgets, and a single device often needs to support multiple sensing applications without deploying a separate model for each application. After deployment, the device continues to operate in changing real-world environments, where shifts in data distribution can degrade the model’s performance over time, so the model must be able to adapt to such shifts efficiently.

The first challenge …

Efficient And Universal Watermarking For Llm-Generated Code Detection, Boquan Li, Zirui Fu, Mengdi Zhang, Peixin Zhang, Jun Sun, Xingmei Wang

Research Collection School Of Computing and Information Systems

Large language models (LLMs) have significantly enhanced the usability of AI-generated code, providing effective assistance to programmers. This advancement also raises ethical and legal concerns, such as academic dishonesty and the generation of malicious code. For accountability, it is imperative to detect whether a piece of code is AI-generated. Watermarking is broadly considered a promising solution and has been successfully applied to identify LLM-generated text. However, existing efforts on code are far from ideal, suffering from limited universality and excessive time and memory consumption. In this work, we propose a plugand- play watermarking approach for AI-generated code detection, named ACW …

Dynamic Spectral Denoising With Global-Context Attention For Multi-Behavior Recommendation, Miaomiao Cai, Yunshan Ma, Fangqi Zhu, Junfeng Fang, Zhijie Zhang, Zhiyong Cheng, Xiang Wang, See-Kiong Ng

Research Collection School Of Computing and Information Systems

Multi-behavior recommendation improves target-behavior predic-tion by exploiting heterogeneous auxiliary feedback (e.g., view,collect, and cart), yet its robustness is often undermined by behavior-dependent noise and inconsistency. We argue that the key bottle-neck is not merely noisy behaviors, but a representation-level failurecaused by two coupled heterogeneities. First, intra-behavior rep-resentation entanglement arises when multi-hop propagationblends incidental signals with true preferences in the embeddingspace. This entanglement renders coarse spatial denoising inef-fective, since it cannot suppress noise without sacrificing weak-but-informative niche signals. Second, inter-behavior reliabilityheterogeneity complicates cross-behavior fusion, as the predic-tive value of auxiliary behaviors varies substantially across usersand contexts. Without reliability calibration, aggregation can …

Hvi-Cidnet+: Beyond Extreme Darkness For Low-Light Image Enhancement, Kangbiao Shi, Xiaowen Ma, Yixu Feng, Tao Hu, Peng Wu, Guansong Pang, Qingsen Yan

Research Collection School Of Computing and Information Systems

Low-Light Image Enhancement (LLIE) aims to recover visually pleasing content and details from degraded low-light images. However, existing RGB-based methods often suffer from color bias and brightness artifacts due to inherent high color sensitivity. Although the HSV color space can decouple brightness and color, it introduces noticeable red and black noise artifacts. To address these challenges, we adopt the Horizontal/Vertical-Intensity (HVI) color space for LLIE, which is defined by the HV color map and learnable intensity. The former enforces small distances for red coordinates to alleviate red noise artifacts, while the latter adaptively compresses low-light regions to suppress black noise …

Audeter: A Large-Scale Dataset For Deepfake Audio Detection In Open Worlds, Qizhou Wang, Hanxun Huang, Guansong Pang, Sarah Erfani, Christopher Leckie

Research Collection School Of Computing and Information Systems

Speech synthesis systems can now produce highly realistic vocalisations that pose significant authenticity challenges. Despite substantial progress in deepfake detection models, their real-world effectiveness is often undermined by evolving distribution shifts between training and test data, driven by the complexity of human speech and the rapid evolution of synthesis systems. Existing datasets suffer from limited real speech diversity, insufficient coverage of recent synthesis systems, and heterogeneous mixtures of deepfake sources, which hinder systematic evaluation and open-world model training. To address these issues, we introduce AUDETER (AUdio DEepfake TEst Range), a large-scale and highly diverse deepfake audio dataset comprising over 4,500 …

Task-Aligned Haze Removal With Semantic-Aware Fusion And Contrast Self-Correction, Jinbin Wang, Aiping Yang, Guosong Jiang, Wenlong Yu, Dongwei Ren, Qinghua Hu

Research Collection School Of Computing and Information Systems

Adverse haze conditions introduce complex degradations that obscure scene details and distort structural cues critical for object detection, posing persistent challenges for vision‐based sensing systems. Although existing haze removal methods have achieved notable improvements in visual clarity, their optimisation objectives are often misaligned with downstream detection requirements, leading to limited detection performance in real‐world scenarios. To address this issue, this work proposes a task‐aligned weakly supervised haze removal framework, termed Dehaze4Detection, which explicitly aligns low‐level restoration with high‐level detection objectives. The framework incorporates a Semantic‐Aware Multi‐Scale Fusion Module (SMFM) that embeds pixel‐level semantic knowledge into the dehazing process, enabling selective …

Parameterized Approximation Schemes For Fair Clustering In Doubling Metrics, Xiaoliang Wu, Ting Liang, Zhize Li, Qilong Feng

Research Collection School Of Computing and Information Systems

Fair clustering has garnered considerable attention, with various fairness notions proposed to ensure equitable representation across demographic groups. In this paper, we focus on the k-center problem in bounded doubling metrics under two popular fairness requirements: group fairness and data summarization fairness, referred to Group Fair k-Center (Gf-k-Cen) and Data Summarization Fair k-Center (Dsf-k-Cen), respectively. Both fairness notions extend classical clustering formulation by associating each data point with a demographic label. Motivated by recent advances in parameterized approximation results for fair clustering, we investigate whether these problems admit Fixed-Parameter Tractable (FPT) approximation schemes in bounded doubling metrics. The previous algorithms …

Convergence Theory For Deep And Multi-Grade Neural Architectures, Lei Huang

Mathematics & Statistics Theses & Dissertations

This dissertation studies two complementary notions of convergence arising in modern neural network models: the convergence of recursively constructed neural network architectures and the convergence of optimization algorithms used for training neural-network-based image restoration models. The first part develops a convergence theory for deep neural networks (DNNs) viewed as recursively generated sequences of functions. Within an Lp framework motivated by statistical learning, sufficient conditions are established under which increasing-depth neural network sequences converge to well-defined limiting functions. The analysis covers both bounded-width and unbounded-width architectures, establishes explicit convergence rates, and motivates a network initialization strategy derived from the convergence conditions. …

Learning To Unlearn: Unlearning And Meta-Unlearning For Continually Adapting Cybersecurity Threat Detectors, Daniel Lucio

Open Access Theses & Dissertations

Machine learning (ML) models deployed in non-stationary environments must continually adapt to evolving data distributions. This challenge is particularly critical in cybersecurity, where malware, intrusion techniques, and adversarial behaviors evolve over time. Continual learning primarily enables incorporating new knowledge while preserving prior knowledge, however, indiscriminately retaining obsolete and harmful information can hinder future adaptation and consume limited model capacity. We argue that effective adaptation should not only acquire new knowledge, but also selectively discard obsolete and less useful historical knowledge before learning from a new distribution. In this work, we propose a meta-learning framework that learns what to forget to …

Integrated Framework For Tsn-Enabled Ot Networks And Scalable Edge Computing To Enable Real-Time Feedback Loop, Taposh Kumer Sarker

Open Access Theses & Dissertations

The advent of Industry 5.0 envisions smart manufacturing characterized by human centricity, sustainability, and systemic resilience. Realizing this vision requires the seamless convergence of Information Technology (IT) and Operational Technology (OT) networks. However, integrating massive, stochastic IT edge computing workloads with deterministic physical control loops introduces severe architectural friction, inherently threatening the safety guarantees required by industrial machinery. To resolve this fundamental incompatibility, this dissertation proposes the Edge-Augmented Real-Time Industrial Control System (EA-RICS).

EA-RICS is a comprehensive, multi-layered architecture designed to dismantle systemic bottlenecks across the physical data plane, the centralized control plane, and the edge operating system. First, the …

Evaluating Machine Learning Models On Classification Of Novel Cyber Attacks In The Healthcare Domain, Promise Ehimen

Dissertations, Theses, and Projects

The increasing adoption of the Internet of Medical Things (IoMT) has improved healthcare delivery through connected medical devices while simultaneously expanding the cybersecurity risks facing healthcare organizations. Although machine learning based intrusion detection systems have demonstrated high detection accuracy, their ability to respond reliably to previously unseen cyberattacks remains uncertain. This study investigated how a Neural Network model and a Logistic Regression model classified novel cyberattacks within the IoMT environment. The Neural Network and Logistic Regression models were both trained and tested using a subset of the CICIoMT2024 benchmark dataset. The Neural Network achieved 99.82% test accuracy and a 0.94 …

Vision Transformers And Convolutional Neural Networks For Land Use Scene Classification, Arun D. Kulkarni

Computer Science Faculty Publications and Presentations

Land use scene classification (LUSC) from remote sensing imagery plays a critical role in environmental monitoring, urban planning, and sustainable resource management. In recent years, deep learning methods have significantly advanced the state-of-the-art, with Convolutional Neural Networks (CNNs) dominating the field because of their strong ability to capture local spatial features. However, the emergence of Vision Transformers (ViTs) has introduced a new paradigm that models long-range dependencies through self attention mechanisms, potentially enabling improved global context understanding. This study presents a comparative assessment of Vision Transformers and CNN-based architectures for remote sensing land use scene classification. Representative CNN models, such …

A Data-Driven Framework For Mitigating Breast Cancer Overdiagnosis: From Estimation To Risk-Adjusted Computer-Aided Diagnosis, William M. Brown Jr.

LSU Doctoral Dissertations

In Computer-Aided Diagnosis (CAD) of cancer, standard cost metrics (false-positives and false-negatives) fundamentally fail to account for overdiagnosis. Overdiagnosis is a critical scenario where a disease is correctly detected (true-positive) but is biologically indolent and would never have caused the patient harm or symptoms. While widely recognized in the medical community as a major healthcare crisis driving stressful and invasive overtreatment, overdiagnosis remains severely under-researched within computer science and engineering. This dissertation addresses this interdisciplinary gap by defining the three key computational challenges of overdiagnosis: (i) accurate estimation, (ii) harm quantification, and (iii) algorithmic mitigation. To overcome the estimation challenge, …

Ai-Powered Resume Screening, Sang Suh, Numery Zaber

Faculty Publications

Traditional resume screening is manual, slow, and susceptible to bias, and it struggles to keep pace with today’s application volumes. This paper presents a dual-engine, AI-powered resume screening system designed for transparency and reproducibility. The primary (classical) pipeline encodes resumes and job descriptions using Sentence-BERT (SBERT), computes a resume–job match score via cosine similarity, classifies candidates into 25 job categories using XGBoost, and provides model interpretability through SHAP. In parallel, a prompted large language model (LLM) baseline (GPT-4o/4o-mini) outputs a match score and predicted category for comparative analysis. A Streamlit-based interface integrates both engines to support recruiter workflows and human-in-the-loop …