Computer Sciences Commons^™

Empirically Exploring The Physical Realizability Of Adversarial Examples, Ruoyao Wen

McKelvey School of Engineering Graduate Student Theses & Dissertations

The development of autonomous vehicles (AVs) has been accelerated by advancements in deep neural networks (DNNs), which power the complex perception systems necessary for safe and efficient real-world navigation. However, as AVs increasingly integrate into public transportation networks, the robustness of their perception systems against potential vulnerabilities is critical. Among these threats, adversarial attacks—particularly through the use of adversarial patches—pose significant risks. These patches are carefully crafted perturbations designed to mislead DNNs, potentially compromising AV safety by causing incorrect object recognition or misclassification.

While extensive research has demonstrated high attack success rates for adversarial patches in controlled digital environments, their …

Stereotyping In Language (Technologies): An Examination Of Racial And Gender Stereotypes In Natural Language And Language Models, Messi Lee

McKelvey School of Engineering Graduate Student Theses & Dissertations

This dissertation examines stereotyping across natural language and language technologies through three interconnected studies. The first chapter applies a contemporary model of race relations from social psychology to investigate America's racial framework within American English, revealing how language encodes hierarchical associations between racial/ethnic groups and attributes of superiority and Americanness. The second chapter extends this analysis to Large Language Models (LLMs), finding that these language technologies portray socially subordinate groups as more homogeneous compared to dominant groups. The third chapter investigates stereotyping in Vision Language Models (VLMs), showing that these language technologies generate more uniform representations for women than men …

Coding For Decentralized Systems And Forensic 3d Fingerprinting, Canran Wang

McKelvey School of Engineering Graduate Student Theses & Dissertations

This dissertation presents novel coding techniques that optimize communication costs and address security challenges in decentralized systems and 3D printing technologies. The first part focuses on encoding data in distributed systems in a decentralized manner, i.e., without a central processor which orchestrates the operation. In such systems, processors require coded data generated from inputs provided by source processors. An example is a distributed storage system with geographically dispersed nodes storing a large database collected by specific source processors. To reduce communication costs, we propose a universal solution applicable to any linear code, with optimizations for systematic Reed-Solomon and Lagrange codes, …

Multimodal Representation Learning Frameworks For Modeling Progression And Heterogeneity In Alzheimer’S Disease, Sayantan Kumar

McKelvey School of Engineering Graduate Student Theses & Dissertations

Alzheimer’s Disease (AD) is the leading cause of dementia, characterised by cognitive and functional impairments that disrupt daily activities. Different clinical modalities such as neuroimaging biomarkers, cognitive assessments, fluid biomarkers and genetic data provide unique and complementary information, contributing to a more comprehensive understanding of disease progression and heterogeneity in disease characteristics. With recent advancements in computational capabilities, particularly in deep learning, multimodal representation learning frameworks aim to integrate diverse clinical modalities into a cohesive framework, capturing the most significant patterns within each modality. Existing data-driven multimodal representation learning frameworks in AD research have two major limitations. First, AD progresses …

Reimagining The Machine Vision Pipeline In Cyber Physical Systems For Trustworthiness And Efficiency, Adith Boloor

McKelvey School of Engineering Graduate Student Theses & Dissertations

Cyber-physical systems (CPS), including autonomous vehicles, drones, and mobile robots, rely on intricate sensors, actuators, and machine learning algorithms to perceive the physical world and execute actions within their surroundings. In the context of vision-driven CPS, achieving this demands processing a substantial volume of visual data captured by on-board cameras. The data is subsequently channeled through digital processors and harnessed by deep neural networks for tasks such as image classification, object detection, and depth perception. This data-centric, machine-vision-infused CPS fosters intelligent decision-making, thereby enhancing overall system performance. Trustworthiness, encompassing the robustness of the entire machine-vision pipeline, and system-level efficiency are …

Reimagining The Machine Vision Pipeline In Cyber Physical Systems For Trustworthiness And Efficiency, Adith Boloor

McKelvey School of Engineering Graduate Student Theses & Dissertations

Cyber-physical systems (CPS), including autonomous vehicles, drones, and mobile robots, rely on intricate sensors, actuators, and machine learning algorithms to perceive the physical world and execute actions within their surroundings. In the context of vision-driven CPS, achieving this demands processing a substantial volume of visual data captured by on-board cameras. The data is subsequently channeled through digital processors and harnessed by deep neural networks for tasks such as image classification, object detection, and depth perception. This data-centric, machine-vision-infused CPS fosters intelligent decision-making, thereby enhancing overall system performance. Trustworthiness, encompassing the robustness of the entire machine-vision pipeline, and system-level efficiency are …

Multimodal Representation Learning Frameworks For Modeling Progression And Heterogeneity In Alzheimer’S Disease, Sayantan Kumar

McKelvey School of Engineering Graduate Student Theses & Dissertations

Alzheimer’s Disease (AD) is the leading cause of dementia, characterised by cognitive and functional impairments that disrupt daily activities. Different clinical modalities such as neuroimaging biomarkers, cognitive assessments, fluid biomarkers and genetic data provide unique and complementary information, contributing to a more comprehensive understanding of disease progression and heterogeneity in disease characteristics. With recent advancements in computational capabilities, particularly in deep learning, multimodal representation learning frameworks aim to integrate diverse clinical modalities into a cohesive framework, capturing the most significant patterns within each modality. Existing data-driven multimodal representation learning frameworks in AD research have two major limitations. First, AD progresses …

Safety And Optimality Monitors For Learning-Enabled Systems Using Conformal Prediction, Jackson Cox

McKelvey School of Engineering Graduate Student Theses & Dissertations

The use of machine learning to create data-driven plant models and controllers has led to an increased need for safety and optimality monitors for model-based systems. System plant models are subject to uncertainty due to learning constraints such as unseen data and overfitting or physical constraints such as unknown dynamics and noise. This uncertainty is detrimental to safety-critical systems and must be properly regulated. To curb this uncertainty, we create prediction sets using the guarantees provided by Conformal Prediction. With a user-specified high probability, these prediction sets contain the true plant system states for an entire prediction horizon, which we …

Schedulability Analysis Of Multi-Phase Limited-Preemption Tasks, Benjamin Standaert

McKelvey School of Engineering Graduate Student Theses & Dissertations

This work addresses hard real-time systems, in which tasks must be scheduled so that they are guaranteed to meet deadlines. In particular, when tasks execute across multiple domains with high preemption costs, the combined cost of these preemptions can cause the system to become unschedulable. The number of preemptions must therefore be bounded to limit the overall task execution time, while ensuring that task blocking times are small enough to allow the system to be schedulable. Prior work introduces the Multi-Phase Secure model, which describes a more exact version of this scenario, and an algorithm to determine schedulability of sporadic …

Improving Clinical Information Extraction From Electronic Health Records: Leveraging Large Language Models And Evaluating Their Outputs, Kriti Bhattarai

McKelvey School of Engineering Graduate Student Theses & Dissertations

Accurate extraction of clinical entities and phenotypes from unstructured electronic health record (EHR) text is crucial for various clinical research tasks, including cohort identification, tracking temporal patterns in disease progression and deciding treatment course. However, this task remains challenging due to the complexity and ambiguity of medical language. This dissertation explores the application of advanced generative pre-trained transformer (GPT) models, such as GPT-4, GPT-3.5-turbo, Llama-3.1, Llama-3 and Flan-T5, for clinical entity and phenotype extraction from EHRs. Building upon these findings, this dissertation also investigates a hybrid approach where integration of external knowledge sources, such as Unified Medical Language System (UMLS) …

Internet Of Medical Things: Integrating Machine Learning And Wearables For Healthcare, Jingwen Zhang

McKelvey School of Engineering Graduate Student Theses & Dissertations

The Internet of Medical Things (IoMT), which integrates Internet of Things technologies into healthcare, has become a powerful tool for enhancing health monitoring and predicting clinical outcomes. By leveraging wearable devices, IoMT facilitates continuous, cost-effective, and convenient tracking of patients' health conditions over time. This dissertation applies data-driven methods to address critical clinical challenges involving wearable devices. Specifically, it focuses on three significant clinical problems: (1) indoor contact tracing for healthcare workers using Bluetooth Low Energy (BLE) beacons, (2) predicting surgical outcomes using wearable data, and (3) developing robust models for surgical outcome prediction that account for patient variability using …

Internet Of Medical Things: Integrating Machine Learning And Wearables For Healthcare, Jingwen Zhang

McKelvey School of Engineering Graduate Student Theses & Dissertations

The Internet of Medical Things (IoMT), which integrates Internet of Things technologies into healthcare, has become a powerful tool for enhancing health monitoring and predicting clinical outcomes. By leveraging wearable devices, IoMT facilitates continuous, cost-effective, and convenient tracking of patients' health conditions over time. This dissertation applies data-driven methods to address critical clinical challenges involving wearable devices. Specifically, it focuses on three significant clinical problems: (1) indoor contact tracing for healthcare workers using Bluetooth Low Energy (BLE) beacons, (2) predicting surgical outcomes using wearable data, and (3) developing robust models for surgical outcome prediction that account for patient variability using …

A Limited-Preemption Scheduling Model Inspired By Security Considerations, Benjamin Standaert, Fatima Raadia, Marion Sudvarg, Sanjoy Baruah, Thidapat Chantem, Nathan Fisher, Christopher Gill

Computer Science and Engineering Faculty Research

Safety-critical embedded systems such as autonomous vehicles typically have only very limited computational capabilities on board that must be carefully managed to provide required enhanced functionalities. As these systems become more complex and inter-connected, some parts may need to be secured to prevent unauthorized access, or isolated to ensure correctness.

We propose the multi-phase secure (MPS) task model as a natural extension of the widely used sporadic task model for modeling both the timing and the security (and isolation) requirements for such systems. Under MPS, task phases reflect execution using different security mechanisms which each have associated execution time costs …

Role Of Vegfa In Intervertebral Disc Pathoanatomy And Low Back Pain, Ryan Potter

McKelvey School of Engineering Graduate Student Theses & Dissertations

Chronic low back pain (LBP) affects more than 80% of Americans and causes significant disability and staggering public health costs. Despite this astonishing prevalence, there are currently no disease modifying therapies. Intervertebral disc (IVD) degeneration accounts for a considerable proportion of LBP, but the precise mechanisms driving the painful pathoanatomy are not understood. The degenerating IVD exhibits a complex array of inflammatory and physiologic changes, and many are associated with the presentation of LBP. For example, the presence of nerves and blood vessels in degenerate IVDs have been identified in patients and animals presenting with LBP symptoms. However, no studies …

Learning And Planning In Stochastic Interactive Environments With The Presence Of Sparse Dependence Structures, Zihao Deng

McKelvey School of Engineering Graduate Student Theses & Dissertations

When interacting with an environment, an agent would want to make decisions on the fly based on the input of the dynamically changing environmental factors, so as to maximize the rewards or minimize the risks. In order to achieve that, one would want to learn the dynamics of the environment and employ efficient planning based on what has been learned about the environment. However, when the environment is large and convoluted, the planning and learning algorithms can easily become computationally intractable in general. In this thesis, we first quantitatively show that this type of problems is indeed hard to solve …

Origin, Regulation, And Function Of Bone Marrow Adipose Tissue And Implications For Bone Health, Xiao Zhang

McKelvey School of Engineering Graduate Student Theses & Dissertations

Bone marrow adipose tissue (BMAT) is a unique fat depot located within the skeletal system that takes up a large portion of the total bone marrow volume and contains tremendous amounts of energy that can be potentially utilized to fuel the body. However, largely attributed to its strong resistance to lipolytic stimuli and its persistent accumulation in various physiological and pathological conditions, the exact function of BMAT within the bone and how it is regulated throughout the body remains largely unclear. This dissertation sought to better understand the unique role of BMAT within the bone marrow niche by first reviewing …

Application Of Machine Learning For Multi-Omics Data Integration, Dhoha Abid

McKelvey School of Engineering Graduate Student Theses & Dissertations

Traditionally, machine learning (ML) is used to train a model to predict scores for instances that were not seen by the model during training. In this conventional use of ML, the model is trained to learn general patterns that relate features to labels, so it can predict accurate scores for unseen data. Here, we use ML, unconventionally, to integrate different types of noisy data. Specifically, for biological investigations, in which there is no mean to measure a ground truth. We propose to train a ML model to predict scores on the same instances that were used in its training. In …

Projection-Domain Low-Count Quantitative Spect (Lc-Qspect) Methods For Radiopharmaceutical Therapies (Rpts), Zekun Li

McKelvey School of Engineering Graduate Student Theses & Dissertations

Radiopharmaceutical therapies (RPTs) using α- or β-particle-emitting isotopes are becoming increasingly important in cancer treatment. Reliable (accurate and precise) quantification of absorbed doses in lesions and vital organs is important for the safety and effectiveness of these therapies. Quantitative single-photon emission computed tomography (SPECT) provides a mechanism for such dose quantifications by quantifying the regional activity uptake. This dissertation aims to develop methodologies for reliable SPECT-based regional uptake quantification for patients treated with α-particle-emitting RPTs (α-RPTs). A significant challenge with conventional reconstruction-based quantitative (RBQ) SPECT methods is the ill-posed nature of image reconstruction, which is further complicated by the limited …

Experimental Design For Scientific Discovery, Quan Minh Nguyen

McKelvey School of Engineering Graduate Student Theses & Dissertations

Experimental design offers an elegant model of many problems where one navigates within a vast search space seeking data points with certain characteristics. A multitude of applications in science and engineering fall under this umbrella, with drug and materials discovery being prime examples. The experimental design approach maintains a probabilistic model of the search space, and uses Bayesian decision theory accounting for this model to guide the accumulation of observed data to maximize an experimentation objective of interest. This dissertation explores Bayesian optimization and active search, two realizations of the experimental design framework that model discovery tasks. While existing solutions …

Tree Recovery By Dynamic Programming, Gustavo Alberto Gratacos

McKelvey School of Engineering Graduate Student Theses & Dissertations

Tree-like structures are common, naturally occurring objects that are of interest to many fields of study, such as plant science and biomedicine. Analysis of these structures is typically based on skeletons extracted from captured data, which often contain spurious segments or cycles that need to be removed. We propose a dynamic programming algorithm which seeks to recover directed trees from these noisy skeletons. Our method recovers trees by removing edges and duplicating nodes while adhering to edge-label constraints. Our algorithm proceeds by iteratively merging graph nodes, such that the solution on the original graph can be obtained from those on …

Improved Models Of Elastic Scheduling, Marion Baumli Sudvarg

McKelvey School of Engineering Graduate Student Theses & Dissertations

In real-time computing systems, \textit{timely} execution is a requirement of \textit{correct} execution. Such systems are widely found in robotics and autonomous vehicle applications, mobile spectrometry of atmospheric aerosols, real-time hybrid simulation for natural hazards engineering, and in prompt localization of transients such as gamma-ray bursts for time-domain and multi-messenger astrophysics. \textit{Elastic scheduling} provides a framework to adjust computational rates and workloads in systems for which timeliness cannot otherwise be guaranteed. While originally proposed for periodic tasks executing on a single processor, elastic scheduling has since been extended to sequential and parallel execution on multiple processors and to earliest deadline first …

Mending Trust In Ai: Trust Repair Policy Interventions For Large Language Models In Visual Data Journalism, Hangxiao Zhu

McKelvey School of Engineering Graduate Student Theses & Dissertations

Trust in Large Language Models (LLMs) emerged as a pivotal concern. This is because, despite the transformative potential of LLMs in enhancing the interpretability and interactivity of complex datasets, the opacity of these models and instances of inaccuracies or biases have led to a significant trust deficit among end-users. Moreover, there is a tendency for people to personify AI tools that utilize these LLMs, attributing abilities and sensibilities that they do not truly possess. This thesis exploits this personification and proposes a comprehensive framework of trust repair policies tailored to address the challenges inherent in LLM annotations within data journalism …

Capturing Higher-Order Relationships Through Information Decomposition, Aobo Lyu

McKelvey School of Engineering Graduate Student Theses & Dissertations

Mutual information between two random variables is a well-studied notion, whose understanding is fairly complete. Mutual information between one random variable and a pair of other random variables, however, is a far more involved notion. Specifically, Shannon's mutual information does not capture fine-grained interactions between those three variables, resulting in limited insights in complex systems. To capture these fine-grained higher-order interactions among variables, Williams and Beer proposed a framework called Partial Information Decomposition (PID) to decompose this mutual information to information atoms, called unique, redundant, and synergistic, and proposed several operational axioms that these atoms must satisfy. This conceptual …

Interpretable Deep Learning Via Sparse Representation For Protein-Dna Interactions, Shane Kuei-Hsien Chu

McKelvey School of Engineering Graduate Student Theses & Dissertations

A central theme in modeling in regulatory genomics is to consider what representation is suited for DNA sequences that give the most helpful information. Traditionally, there are k-mers, position-weight-matrices, parametric statistical models like Hidden Markov models, and, more recently, deep neural network. In this work, we introduce sparse representations as a principled framework for problems in regulatory genomics. We show that sparse representation is a framework that allows us to build techniques to answer challenging inferential questions in regulatory genomics. Leveraging sparse representations, we reveal that gapped and long motifs are prevalent in in-vivo datasets like ChIP-Seq, often corresponding to …

An Assistive Interface For Displaying Novice's Code History, Ruiwei Xiao

McKelvey School of Engineering Graduate Student Theses & Dissertations

As Teaching Assistant (TA) programs grow in number and size in introductory CS courses, TAs play a significant role in novice programmers' experience and contribute to their success. However, many TAs are also relative beginners themselves and thus have limited experience in programming and teaching. Thus the effectiveness and consistency of their guidance can vary significantly. To improve interaction quality and assist TAs in providing better support, we examine the difficulties encountered by inexperienced TAs in previous literature and then identify the potential for the high cognitive load as an unaddressed difficulty that may prevent new TAs from initiating effective …

Feature Selection From Clinical Surveys Using Semantic Textual Similarity, Benjamin Warner

McKelvey School of Engineering Graduate Student Theses & Dissertations

Survey data collected from human subjects can contain a high number of features while having a comparatively low quantity of examples. Machine learning models that attempt to predict outcomes from survey data under these conditions can overfit and result in poor generalizability. One remedy to this issue is feature selection, which attempts to select an optimal subset of features to learn upon. A relatively unexplored source of information in the feature selection process is the usage of textual names of features, which may be semantically indicative of which features are relevant to a target outcome. The relationships between feature names …

Understanding Societal Values Of Chatgpt, Yidan Tang

McKelvey School of Engineering Graduate Student Theses & Dissertations

As Large language models (LLMs) become increasingly pervasive in various domains, it is crucial to ensure that their outputs adhere to societal values and ethical considerations. In this thesis, we investigate the alignment of ChatGPT, a recent state-of-the-art large language model developed by OpenAI, with societal values. Specifically, we define the problem of societal values of LLMs and assemble a representative collection of 7 datasets covering 4 topics related to societal values. In-context learning techniques are applied and appropriate prompts are designed. The performance of each dataset is measured using a standardized evaluation system focused on accuracy. We then display …

Evaluating The Problem Solving Abilities Of Chatgpt, Fankun Zeng

McKelvey School of Engineering Graduate Student Theses & Dissertations

This thesis addresses the need for a fair evaluation of language models' problem solving abilities by presenting a unified evaluation framework for ChatGPT on 16 problem solving datasets (e.g., NaturalQA, HellaSwag, MMLU, etc.). We evaluate the model's performance using F1, exact match, and quasi-exact match metrics and find that ChatGPT is highly accurate in solving tasks that require commonsense and knowledge. However, we also identify truncated text bias and few-shot scenarios as challenges that may impact ChatGPT's performance. Our research highlights the importance of standardizing datasets and developing a unified evaluation system for the fair evaluation of language models. Overall, …

Comments Of The Cordell Institute On Ai Accountability, Neil M. Richards, Woodrow Hartzog, Jordan Francis

Scholarship@WashULaw

These comments are a response to the National Telecommunications and Information Administration's 2023 request for comment on AI accountability (AI Accountability RFC, NTIA–2023–0005).

Responding to NTIA’s recent inquiry into AI assurance and accountability, we offer two main arguments regarding the importance of substantive legal protections. First, a myopic focus on concepts of transparency, bias mitigation, and ethics (for which procedural compliance efforts such as audits, assessments, and certifications are proxies) is insufficient when it comes to the design and implementation of accountable AI systems. We call rules built around transparency and bias mitigation “AI half-measures,” because they provide the appearance …

Speeding Up The Quantification Of Contrast Sensitivity Functions Using Multidimensional Bayesian Active Learning, Shohaib Shaffiey

McKelvey School of Engineering Graduate Student Theses & Dissertations

No abstract provided.