Physical Sciences and Mathematics Commons^™

Hdra-Fusion: Hybrid Detection With Routed Architecture For Manipulation-Aware Ai Face Forgery Detection, Omar Ebeid

Theses and Dissertations

With the rapid advancements in artificial intelligence-based image generation and manipulation tools, it is extremely difficult to detect if an image is genuine or artificially crafted. Despite extensive research in this area, existing image detection systems suffer from three major problems: suboptimal cross-dataset generalization due to shortcut learning of dataset-specific patterns, unreliable probability estimates due to domain shift, particularly in cross-manipulation evaluation settings, and an inability to detect images manipulated by multiple types of manipulations within a single detection framework. To address these limitations, we propose HDRA-Fusion (Hybrid Detection with Routed Architecture), a framework built on the conclusion that different …

Quantifying Noise In Monte Carlo Simulations Of Cosmic Rays In A Shock, Abhro Rahman

Theses and Dissertations

Cosmic rays are a phenomenon known to humankind from the early 1900s, and yet today the details of their physical behavior are not fully settled. While the observed spectral fluxes of cosmic rays follow a power-law trend when plotted against particle energy, the exact mechanisms under which they get to such high energies are still under study. This thesis will undertake a subsection of that task, where the Monte Carlo simulation of particle accelerations is described in Warren (2015), and we aim to identify and quantify the noise inherent in such simulations. This document takes a long path through particle …

Hybrid Machine Learning For Zero-Day Malware Detection: An Adaptive Static-Dynamic Analysis Approach, Garrett Farmer

Theses and Dissertations

In an era of swiftly evolving cyber threats, zero-day malware continues to be one of the most challenging classes of attacks to detect and mitigate. Traditional signature-based methods often fail to detect novel malicious code, leaving institutions vulnerable to unknown exploits. This thesis proposes a machine learning (ML)-based framework that is designed to detect unknown malware variants. By combining both static and dynamic techniques, such as file structure exam ination and sandbox-based runtime analysis, this approach aims to successfully capture malicious characteristics. The proposed custom pipeline addresses the computational overhead that is associ ated with deep inspections, outlining a staged …

Bounding The Average Kissing Number, Including New Bounds For Three-Dimensional Binary Sphere Packings, Mark William Bockhaus

Theses and Dissertations

The average kissing number is defined as the supremum over all sphere packings of the value: two times the number of tangencies divided by the number balls in the packing. In this paper, we present a survey of the literature about bounding the average kissing number, beginning with the first non-trivial results, through the most up-to-date bounds. We then turn our focus to binary sphere packings: those which contain spheres of two different radii. We improve upon known bounds for the average kissing number for binary sphere packings in three-dimensions and find exact bounds for many packings.

Modeling Causal Interactions Across Brain Functional Systems For Population-Specific Disease Analysis, Alissen Moreno

Theses and Dissertations

Functional brain connectivity provides critical insight into neural mechanisms underlying neurodegenerative and affective disorders. Traditional neuroimaging studies rely on undirected or region-specific connectivity measures developed predominantly using racially homogeneous cohorts, limiting fairness and generalizability across diverse populations.

We propose a population-aware framework modeling directed causal interactions across large-scale brain functional subnetworks. Resting-state fMRI data from the HABS-HD cohort were used to construct subject-level causal connectivity matrices via ICA-LiNGAM, aggregated into interpretable system-level hyper-connectomes representing interactions among eleven canonical brain subsystems. These features trained nonlinear models for Alzheimer’s disease stage classification and trait worry prediction.

Results demonstrate that MLP models outperform …

Vision-Language Guided Quadruped Navigation With Reinforcement Learning Control, Chen Yuan Wang

Theses and Dissertations

Large Language Models and their multimodal variants, Vision-Language Models (VLMs), can interpret natural language instructions and visual context. Reinforcement Learning (RL) methods such as Proximal Policy Optimization (PPO) have been shown to successfully learn stable quadruped locomotion. This work investigates whether a VLM integrated with an RL controller can control a quadruped robot to perform instruction-following navigation to a predefined target using text instructions, egocentric vision, and proprioceptive state.

Most VLM-based robotics work directly target prediction of robot actions from multimodal inputs. This approach often depends on large amounts of expert demonstration data, generated by controllers such as trained RL …

On The Structure Of The Homotopy Lie Algebra Of Local Rings, Dawson M. Strong

Theses and Dissertations

This thesis investigates the construction and homological properties of the homotopy Lie algebra π(R) of a commutative local ring (R,m,k). Drawing upon the theoretical framework of differential graded (DG) algebras, we first establish the theory of minimal free resolutions and other standard topics in homological algebra. The core of this work details the iterative construction of the acyclic closure R⟨Y⟩ of k over R, which is achieved by the systematic adjunction of exterior and divided power variables to eliminate cycles in homology. We demonstrate that this acyclic closure serves as a minimal free resolution and provides the means to define …

Accurate Polyp Segmentation With Visual Mamba, Diego Adame

Theses and Dissertations

Medical image segmentation is a fundamental task in computer-aided diagnosis because it enables precise delineation of anatomical structures and pathological regions from clinical images. While convolutional neural networks and Transformer-based models have achieved strong performance on many medical segmentation benchmarks, CNNs struggle to model long-range dependencies and Transformers often incur high computational complexity for high-resolution medical images. Recently, state space models have emerged as an efficient alternative for dense prediction tasks due to their ability to capture long-range dependencies with linear complexity. However, existing Mamba-based segmentation models still rely on pixel-wise raster scanning that disrupts spatial locality and use simple …

Mcgehee Blow Up And Collision Manifold Of Planar (2+2)-Body Problem, Nathaniel Scott Sill

Theses and Dissertations

This thesis analyzes the dynamics of the Planar (2+2)-Body Problem which consists of two asteroids moving under the gravitational force of each other and of two larger primaries. We use the McGehee blow up technique to remove the singularities in the dynamics associated with a triple collision with one of the primaries by introducing a new set of variables. Additional variables are introduced to reduce the dimension of the problem. We then derive the dynamics for these new variables. We then use the energy relation that comes from the original Hamiltonian to describe the collision manifold which is pasted in …

Differentiable Objectives For 3d Scene Relighting Via Gradient Descent On Olat Basis Coefficients, Anson Savage

Theses and Dissertations

Designing effective lighting is an iterative and often time-consuming process. This work contributes to automatic lighting design research by presenting a render-engine agnostic optimization routine: gradient descent on RGB multipliers of one-light-at-a-time (OLAT) basis images. We compare several objective functions to accomplish lighting tasks and show that our method is capable of quickly and effectively exploring different lighting styles using either text prompts or reference images. We also present several datasets specific to lighting tasks and show that fine-tuning on these datasets can improve performance.

Maximizing Uinta Basin Oil Well Yield Through Geochemical Fingerprinting Of Produced Water, Troy W. Day

Theses and Dissertations

The Uinta Basin is an unconventional oil play that has, since 2017, annually increased hydrocarbon extraction through modern drilling and fracing techniques with increased frac loading. The primary production target is the Uteland Butte Member of the Green River Formation, which consists of interbedded limestones, dolostones, and thick, organic-rich mudstones. Along with increased hydrocarbon production, volumes of produced water from the Uteland Butte Member have also risen. Water is co-produced with hydrocarbons and is an undesirable byproduct of hydrocarbon extraction. Large frac operations now consistently have highly variable percentages of recovered water volume relative to oil. This increase in water …

Probing The Disequilibrium Chemistry Of Two Late T-Dwarfs Using Planetary Intensity Code For Atmospheric Scattering, Crystal-Lynn Jacqueline Bartier

Theses and Dissertations

We have performed a detailed analysis of the atmospheric properties of two late-T type brown dwarfs, 2MASSI J0415195-093506 and 2MASS J05591914-140448. For each brown dwarfs, we combined the spectra of each brown dwarf from the ground at 1-2.5 µm, AKARI at 2.5-5 µm, and from the Spitzer Space Telescope at 5-15 µm. For this project, we are particularly interested in probing the extent of vertical mixing. For this analysis, we used the Planetary Intensity Code for Atmospheric Scattering (PICASO), which is a code that creates a radiative transfer model for a given set of atmospheric parameters. With PICASO we find …

A Masked-Perturbation Model For Destabilization Security Games In Feedback Systems, Gabriel Michael Perry

Theses and Dissertations

In this thesis, we present a mathematical formulation for increasing cyberphysical system security. This model allows a system administrator to use assumptions about exposed signals to derive defense strategies meant to mitigate the system's vulnerability to feedback perturbation attacks. We begin with a survey of game theory with the goal of building the concepts of resource and security games. We then introduce our model, which is a security game; targets in this game are exposed signals in the model map of the system, and resources are perturbed or introduced dynamical relationships between these signals. This formulation, called the masked-perturbation model, …

Theoretical Expressivity Of Self-Attention As A Model Of Binary Sequence-To-Sequence Functions And Other Explorations, Dustin Angerhofer

Theses and Dissertations

The self-attention mechanism lies at the heart of modern sequential modeling with deep neural networks, especially in the context of Large Language Models (LLMs). The basic role of self-attention is to capture pairwise interactions between tokens in a sequence. This allows token embeddings to be endowed with context-dependent information before further processing. Despite its widespread use, our theoretical understanding of self-attention is still nascent. In this work, we examine self-attention as a model of binary sequence-to-sequence (seq2seq) functions, which is the simplest non-trivial case to analyze. We find, using VC dimension and the Sauer-Shelah lemma, that in the limit of …

The Lookahead Transformer, Justin P. Olcott

Theses and Dissertations

Large Language Models (LLMs) are transformer models that generate text autoregressively in two stages: prefill and decode. The prefill stage processes all input tokens in parallel and caches them into the KV cache. The decode stage uses the KV cache, processes a single token at a time, and is generally memory-bandwidth constrained. Each decode step generates a token probability distribution, samples from it, appends the token to the cache, and repeats. This left-to-right, one-at-a-time generation paradigm has proven effective for language modeling in many domains; however, it also has limitations such as token permanence, constant compute per token generation, constant …

Detection Of The Presence And Growth Of E. Coli Bacteria In A Droplet-Based Microfluidic System Using Laser Induced Fluorescence, Oshira Peiris

Theses and Dissertations

Antimicrobial resistance is a biomedical event, wherein an infectious microbe becomes non-responsive to the dosage of a particular antibiotic due to previous overuse or overexposure. Currently, agar plate disc diffusion is used as the gold standard for detecting antimicrobial resistance. The possibility of miniaturizing and speeding up detection techniques for antimicrobial resistance is investigated in this thesis. Microfabrication techniques were used in developing two droplet microfluidic systems interfaced with laser induced fluorescence detection. Microfluidic devices were manufactured with different droplet generating geometries using poly-dimethyl siloxane, soft lithography and hydrophobic surface treatment. The microfluidic devices were tested for droplet quality using …

Model Unification: A Systematic Approach For The Cluster Expansion, Kolten Barfuss

Theses and Dissertations

A systematic approach for model unification of cluster expansion models is developed and demonstrated. The method builds on previous work with the Manifold Boundary Approximation Method (MBAM) to identify supremal models. The work is motivated by the exercise of unifying cluster expansion models – characterized by Π matrices – for two binary alloys that share one atomic species to form a model for the ternary alloy. The result is the Supremal Π Algorithm, a systematic unification algorithm that relies on an MBAM-informed understanding of models and on the singular value decomposition. The algorithm yields ΠΛ, a Π matrix for the …

Some Closure On The Relation Between Cfg And Peg, Nathan J. Bulloch

Theses and Dissertations

We work toward a practical, unified language formalism for structurally recursive languages by expressing the mathematical theories underlying context-free grammars with lookahead (CFGLA) in the Rocq theorem prover and leveraging CFGLA semantics to analyze interactions between between recursion and negative lookahead. We give examples of grammars with left-recursive negation cycles for the parsing expression grammar (PEG) variant due to Medeiros et al. and for the translation from PEG to CFGLA due to Miyazaki and Minamide. To study the interaction between left recursion and negative lookahead in a PEG context, we introduce FPEG, a left-recursive extension of PEG with Fitting semantics.

Asymptotic Sensitivity And Adjoint Methods For Equation Discovery With Open-Source Implementation, Nathan Schill

Theses and Dissertations

The "on-the-fly" (OTF) method is a gradient-based approach to equation discovery. It makes use of (potentially limited) observational data, a data assimilation technique such as continuous data assimilation (also called nudging), and an error metric. From these it computes an asymptotic approximation of a simulated system's error sensitivity with respect to unknown parameters and utilizes any gradient-based optimization algorithm to perform parameter updates. These parameter updates may be done online or offline, and the model may be of a precisely specified form or simply a library of possible terms with potentially nonlinear parameters. We have applied the adjoint method to …

Sonic Booms From Rockets: Measurements And Modeling, Mark Charles Anderson

Theses and Dissertations

Rocket sonic booms are produced when launch vehicles or boosters travel faster than the local speed of sound in the atmosphere, generating systems of conical shock waves. The purpose of this dissertation is to perform high-fidelity measurements of rocket sonic booms and to explain the physics behind the unique phenomena observed. After a thorough literature review on sonic booms produced during spaceflight operations, Part I of this dissertation deals with measurement and analysis challenges. One challenge is reducing contaminating noise effects on sonic boom perception metrics, addressed using a low-pass filtering method tied to the noise floor. A second challenge …

Unlocking Inexpensive Training Through Overcoming Weaknesses In Dataset Distillation, Connor Wilhelm

Theses and Dissertations

Dataset distillation is a nascent technique that promises accelerated neural network training. In particular, distillation produces a highly compressed dataset that can train a neural network in as little as a single optimization step. However, distillation has not been adopted for real-world use cases due to weaknesses, such as poor approximation of the original task, unclear interpretability, high costs of the initial distillation, and poorly elaborated use cases. Here, we present five papers that advance task distillation towards real-world adoption. The papers focus on: distilling reinforcement learning environments, interpreting distillation-produced datasets, improving distillation performance using ensembling, examining distillation's generalization to …

Understanding The Role Of Tiktok As A Source Of Support For Cyberbullying Victims, Saba Iqbal

Theses and Dissertations

Cyberbullying poses a significant threat to teenagers and young adults, who often seek support and advice online. Little is known about the support and advice shared online for victims of cyberbullying and whether it is helpful. In this study, we used qualitative thematic analysis to examine a sample of 150 TikTok videos that give support or advice for cyberbullying victims. We identify 33 unique pieces of support and advice across 5 categories, with differences in tone and delivery among different types of videos. We then evaluated the quality of 33 unique pieces of cyberbullying advice in a user study with …

Advancing Cyclic Ion Mobility Mass Spectrometry For Structural Characterization, Isomer Discrimination And Gas-Phase Ion Chemistry, Sudam Shivaji Mane

Theses and Dissertations

This dissertation focuses on the development and application of high-resolution ion mobility-mass spectrometry (IMS-MS) methodologies for gas-phase separation and structural characterization of small molecules, with particular emphasis on epimers, positional isomers, and protomers. In addition to structural separations, this work investigates fundamental aspects of gas-phase ion chemistry using IMS-MS in combination with collision-induced ion activation and dissociation techniques. Advanced separations were performed using cyclic traveling-wave ion mobility spectrometry (cyclic TWIMS), implemented in a commercial cyclic ion mobility-mass spectrometry (cIMS) platform, to systematically evaluate both the capabilities and limitations of high-resolution IMS in resolving subtle structural differences. Initial studies demonstrated that, …

Practical, Scalable, And Cost-Effective Nanolc-Ms Workflow For Single-Cell Proteomics, Siqi Huang

Theses and Dissertations

Single-cell proteomics provides a novel research strategy for decoding cellular heterogeneity, revealing the molecular mechanisms of rare cell subpopulations, and studying the dynamic responses of cells to external stimuli. However, compared to transcriptomics, single-cell proteomics faces even more challenging technical challenges: extremely low protein amount in single cells, wide dynamic range, significant adsorptive losses during sample preparation, and the complex matrix and trace sample injection demanding extremely high stability and sensitivity from the separation and detection systems. Mass spectrometry-based bottom-up proteomics has become an important tool for analyzing complex biological systems, and nano-flow liquid chromatography-mass spectrometry (nanoLC-MS) exhibits unique advantages …

Differentials As Models: Constructing A Differentials-Within-Limits Framework For First-Year Calculus And Examining Students' Understanding Of Limits, Differentials, And Derivatives Within It, Alexander A. Swindler

Theses and Dissertations

Work in first-year calculus has clearly shown the power that treating differentials (such as dx) as tiny quantities affords for first-year calculus learning, yet a traditional limits-based approach leaves no room for them. While some researchers have loosely described informal differentials and limits within the same paradigm, their actual conceptual definitions and relationships have not been explicitly constructed. Continued progress in this area could benefit from theoretical consideration of (a) providing clearer intuitive definitions for differentials inside a limits-based approach to calculus, and (b) defining more exactly the relationships between differentials and limits. In this study, I propose a theoretical …

An Efficient Approximation To The Vibrational Coupled Cluster Method For Large Molecules, Nivedhitha P Ms

Theses and Dissertations

The Watson Hamiltonian is a widely used model for describing anharmonic vibrational motion in polyatomic molecules. It expresses the kinetic energy in separable dimensionless normal coordinates and represents the potential energy as a Taylor series expansion around the equilibrium geometry, usually up to the quartic approximation. Several computational methods have been developed to solve the corresponding vibrational Schrödinger equation for predicting infrared (IR), Raman, and vibrational circular dichroism (VCD) spectra. These include the Vibrational Self-Consistent Field (VSCF) method, Vibrational Configuration Interaction (VCI) and its approximations, and Vibrational Second-Order Perturbation Theory (VPT2), each providing different balances between computational cost and accuracy. …

Bridging The Cybersecurity Education Gap: The Role Of Open Educational Resources In Supporting Rural Cybersecurity Programs, Brittni Hardie

Theses and Dissertations

This study examines the intersection of cybersecurity education, open educational resources (OER), and rural higher education through a systematic review of current literature and an exploratory survey of rural community college faculty. The purpose of this research, consistent with the approved Institutional Review Board (IRB) protocol, was to understand how OER can be leveraged to design and deliver an affordable, high-quality System Security course within a rural higher-education environment. As cybersecurity workforce shortages continue to grow across the United States, rural institutions face persistent challenges in sustaining high-quality programs due to financial constraints, limited faculty capacity, and rapidly evolving curriculum …

Empirical Benchmarks For Interpreting Effect Sizes In Violent Crime Interventions, Kohta James Matsukawa Hansen

Theses and Dissertations

Criminal justice researchers apply Cohen’s (1988) benchmarks to classify effect sizes as small, medium, or large, despite these standards never being meant for broad, decontextualized use (Cohen, 1988; Gies et al., 2024; Goulet-Pelletier & Cousineau, 2018; Lakens, 2013; Milner et al., 2023). Repeatedly doing so may weaken statistical validity, distort findings, and impede effective policymaking. This study introduces the first effect size benchmarks specifically designed for violent crime interventions.

Using a quasi-meta-analysis framework, 1,605 effect sizes from 104 violent crime intervention studies from the CrimeSolutions clearinghouse were converted to Cohen’s d. Three new discrete benchmarking methods were created using a …

Factorized Electron-Nuclear Dynamics: Theory Development And Implementation, Julian Stetzler

Theses and Dissertations

In the Born-Oppenheimer (BO) approximation, nuclei are treated as classical particles. The theoretical and computational approaches that go beyond it are impractical for all but the smallest chemical systems. However, such methods are necessary to simulate many photochemical and biological processes. This thesis discusses two important topics of research in this feld: frst, improving electronic structure theory to efciently describe quantum nuclei; second, developing a scalable quantum dynamics method, the factorized electron-nuclear dynamics (FENDy).

Following the BO approximation, the electronic structure of molecules is routinely examined for static arrangements of point nuclei, which is a highly useful premise but not …

Examining Marine Fungal Nitrous Oxide Production, Qintiantian Nong

Theses and Dissertations

Nitrous oxide (N2O), a potent greenhouse gas, can be produced by microorganisms inhabiting marine environments. While bacterial denitrification is a major source of N2O, the role of marine fungi in this process remains poorly constrained. This thesis investigates the contribution of marine fungi to N2O production from salt marsh sediments, while also isolating new N2O-producing marine fungi and advancing methodological approaches for quantifying denitrification. First, we conducted experiments to quantify the fungal contribution to N2O production using salt and brackish marsh sediments collected from the North Inlet–Winyah Bay National Estuarine Research Reserve in South Carolina during the summer and winter …