Theory and Algorithms Commons^™

Evaluation And Distillation Of Source Code Generation Tasks By Large Language Models, Danny Brahman

Electronic Theses and Dissertations

Large Language Models (LLMs) are predominantly assessed based on their common sense reasoning, language comprehension, and logical reasoning abilities. While models trained in specialized domains like mathematics or coding have demonstrated remarkable advancements in logical reasoning, there remains a significant gap in evaluating their code generation capabilities. Existing benchmark datasets fall short in pinpointing specific strengths and weaknesses, impeding targeted enhancements in models’ reasoning abilities to synthesize code.

To bridge this gap, this thesis introduces two novel contributions: CodeEval and CodeQual. CodeEval is an innovative, pedagogical benchmarking method that mirrors the evaluation processes encountered in academic programming courses. It comprises …

Uniform Stability Of Katyusha In Strongly-Convex Settings, Don Li

University Honors Theses

Acceleration of convergence and reduction of variance constitute a trade-off in the design of stochastic optimization machine learning algorithms. Katyusha was introduced to address this trade-off, synthesizing Nesterov Accelerated Gradient (NAG) and Stochastic Variance-Reduced Gradient (SVRG) into a single first-order optimizer with promising empirical performance. However, the generalization properties of Katyusha remain largely unexplored. We conjecture that, in the smooth quadratic regime (i.e., under assumptions of strong convexity and smoothness of the loss function, and boundedness of gradients), Katyusha is uniformly stable in the sense of Bousquet and Elisseeff. Instantiating our framework for NAG, we extend the use of Lyapunov …

Edge Co-Occurrence Regularization For Node Classification, Kadir Altunel

Theses

We propose a simple yet effective regularization technique for node classification on graphs that leverages edge-based label co-occurrence patterns. We first train an MLP on node features to produce class probability distributions, then compute a fixed penalty matrix from edge-based co-occurrence statistics of these predictions. This penalty matrix, which captures unlikely class combinations on connected nodes, is then used to regularize GNN training without further updates. We evaluate this approach across multiple homophilic datasets (Cora, CiteSeer, PubMed, ogbn-arxiv) and heterophilic benchmarks (Chameleon, Squirrel, Actor, Roman-Empire) using three GNN architectures: GCN, GraphSAGE, and H2GCN. Results show consistent improvements on homophilic graphs, …

Teaching Machines To Deter: Exploring Strategic Deterrence In Ai Models, Will Taylor

Theses/Capstones/Creative Projects

This capstone project investigates whether deterrence can emerge as a meaningful strategy within a zero-sum stochastic game using multi-agent reinforcement learning (MARL). After outlining core concepts in game theory and deterrence, the study models a simplified deterrence environment in which two minimax-Q agents repeatedly interact under uncertainty and adversarial incentives. The agents learn from rewards shaped by escalation costs, unilateral vulnerability, and the stabilizing benefits of restraint. Results show that both agents consistently converge toward a conservative, status-quo strategy, overwhelmingly selecting the Maintain action while avoiding both escalation and restraint in most scenarios. This behavior reflects the risk-averse logic of …

Learning Global Context For Sparse Activity Recognition In Lengthy Recordings With Limited Dataset Size, Zeyu Tang

All Dissertations

This dissertation describes methods to analyze lengthy recordings of data in order to detect sparsely occurring activities. The narrative below describes the progression of research that led to the development of these methods and their generalization into a unified framework. My research started with designing models for dietary monitoring, including detecting meals from day-long recordings and detecting intake gestures from meal-length recordings. Both tasks share some common characteristics: (a) the target event takes only a small portion of data recordings, and (b) there is global context within full-length data recordings that can help a model make better decisions. After finishing …

Dynamic-Query Robustness Of Ann Indexes Under Time-Indexed Drift, Stellamaris Nakacwa, Majid Shaalan

Harrisburg University Other Works

Approximate nearest-neighbor search is a central retrieval primitive in dense question-answering and retrieval-augmented generation systems. Existing ANN evaluation protocols typically measure recall, latency, throughput, and search-effort sensitivity under a fixed-query assumption: a query vector is submitted to an index, approximate neighbors are retrieved, and the result is compared with exact nearest-neighbour ground truth. This assumption is appropriate for conventional vector-search benchmarking, but it is less complete for multi-step, distributed, and agent-controlled retrieval pipelines in which the retrieval-facing query may be refined, recomputed, or displaced across execution steps. This paper introduces a time-driven dynamic query evaluation framework for ANN search. The …

Towards Physics-Informed Neural Networks For Simulating Multiphase Geothermal Convection​*, Daniel C. Patton, Andrew Harrison Eno

Campus Research Month

Water and steam flow through porous rock, transferring heat via conduction and buoyancy-driven convection caused by density differences. Traditional numerical methods (finite-volume/finite-element) model this well but can become memory-intensive and unstable for long, high-detail simulations. This work demonstrates a Physics-Informed Neural Network (PINN) using a finite-difference approach within the NVIDIA PhysicsNeMo framework to simulate magma chambers in 2D. Tested on the Rio Pisco pluton in Peru, results are compared with the USGS HYDROTHERM model. PINNs learn from physical laws, offering accurate, flexible solutions with less data and development effort.

Scenarioxp: A Complete Scenario-Based Testing Framework For The Exploration And Exploitation Of Autonomous Vehicle Validation Scenarios, Quentin Goss

Doctoral Dissertations and Master's Theses

Today is an age of exciting emerging technology where cutting-edge research in autonomous vehicles (AVs) reduces the active human participation in driving and extends awareness beyond human limitations of perception and reaction, improving driving safety and quality of the user experience as a result. The ever-increasing complexity of these autonomous systems poses many challenges towards the validation and verification (V\&V) of these complex systems under time and resource constraints, as the use of artificial intelligence and also the intricacy of the operating environment means that these systems are also black-box and non-deterministic. Scenario-based V\&V testing of such systems, which involves …

Study Of Output And Behavior Of Llms Using Confidence Framing In Prompt Engineering, Micah Parrilla

Doctoral Dissertations and Master's Theses

While prompt engineering is pivotal for shaping Large Language Model (LLM) outputs, the impact of confidence framing on behavioral calibration remains underexplored. This study investigates the ways in which psychological framing, utilizing techniques such as capability praise, role amplification, and doubt induction, affects linguistic tone, objective accuracy, and internal calibration. A 1,080-trial experimental matrix evaluated six diverse models across factual, logical, coding, and cyber security domains. Analysis using the Kruskal-Wallis H-test revealed highly significant behavioral shifts across all measured dimensions, providing conclusive evidence that the applied frames exert a substantial influence on model performance.

The findings identify a distinct cognitive …

A General Weighting Theory For Ensemble Learning: Beyond Variance Reduction Via Spectral And Geometric Structure, Ernest Fokoue

Articles

Ensemble learning is traditionally justified as a variance-reduction strategy, explaining its strong performance for unstable predictors such as decision trees. This explanation, however, does not account for ensembles constructed from intrinsically stable estimators-including smoothing splines, kernel ridge regression, Gaussian process regression, and other regularized reproducing kernel Hilbert space (RKHS) methods whose variance is already tightly controlled by regularization and spectral shrinkage. This paper develops a general weighting theory for ensemble learning that moves beyond classical variance-reduction arguments. We formalize ensembles as linear operators acting on a hypothesis space and endow the space of weighting sequences with geometric and spectral constraints. …

Addressing The Problems Of Data Variations, Quality, And Scarcity In Training Deep Neural Networks, Jian Sun

Electronic Theses and Dissertations

The performance of deep neural networks (DNNs) is strongly influenced by the characteristics and quality of the underlying datasets. This Ph.D. dissertation addresses three pervasive data challenges-imbalance, quality degradation, and scarcity-that commonly hinder the effectiveness of DNNs in computer vision (CV) and natural language processing (NLP) applications.

Class imbalance remains one of the most frequent causes of degraded model generalization. While Focal Loss effectively mitigates inter-class imbalance by assigning higher weights to minority classes, it struggles with intra-class imbalance, particularly in video datasets where longer clips dominate feature representation. To address this, I implement and utilize …

Evolving Solutions For Red Blood Cell Preservation, Ali Alkafaji, Charles A. Elder, Mohammad Zaidi, Kavin Parthiv, Michael A. Menze

The Cardinal Edge

In emergencies such as natural disasters, armed conflicts, or during outer space missions, the availability of transfusable blood can mean the difference between life and death. Red blood cells (RBCs) must be stored at +4 ± 2 °C and have a shelf life of just 42 days, which makes maintaining a stable blood supply during adverse conditions extraordinarily challenging. This challenge was especially apparent during the COVID-19 pandemic when hospitals faced severe blood shortages. Freeze-drying, or lyophilization, offers a promising avenue to extend the shelf life of RBCs for transfusion during crises. However, a significant hurdle in dry preservation is …

Cerebral Documents And Algorithmic Sensemaking: Searching For Expressions In Human And Artificial Cognitive Collaborations, Rebekah L. Cowell

Proceedings from the Document Academy

Generative Artificial Intelligences (AIs) and current advanced large language models (LLMs) are algorithmically designed to generate text-based conversations as conversational agents (CAs), by replicating human language and conversational communication. Pairing human cognition with generative computationally coded cognition. We have never been here before: cerebral and artificial information collaborations and processing producing expressions that may or may not become visible as second-hand/secondary source documents.

Sensemaking or sense(un)making is a unique autonomous human drive cognitively, our information processing is sensemaking in action and expressions and articulations are evidence of the sensemaking cycle. Documentation [expressed or articulated through various mediums] are a product …

Design And Analysis Of Modern Quantum Neural Network Architectures For Intelligent Systems, Lakshmi Chandrakanth Kasireddy, Prabhakara Rao Kapula, Dineshkumar Rajendran, Neha Bharani, Srikanth Pulipeti, Islombek Khushvaktov

Computer Science Faculty Publications

Quantum neural networks (QNNs) offer a principled pathway for integrating quantum computation with machine learning through superposition- and entanglement-based representations. This chapter proposes an architecture-aware design and evaluation framework for modern QNNs, emphasizing robustness and system feasibility alongside predictive performance. Multiple architectures variational QNNs, quantum convolutional neural networks, tensor-network hybrids, and fully quantum models—are assessed under a unified protocol. Experimental analysis shows that the proposed architecture-search–guided QNN achieves 91.8% classification accuracy and an F1-score of 0.914, outperforming fixed-template variational QNNs by approximately 5.6 percentage points. Under depolarizing noise with probability p = 0.10, the proposed model retains 85.3% accuracy, whereas …

Cogram: A Computational Pipeline For Genome Assembly And Reconstruction Using Graph Neural Networks, William Coggins

College of Graduate Studies: Theses & Dissertations

Genome assembly — the reconstruction of a complete DNA sequence from short, overlapping reads — remains a fundamental challenge in computational biology. A central difficulty is distinguishing true genomic overlaps from spurious connections arising from repetitive sequences, a task that traditional assemblers address through hand-tuned heuristic rules applied to de Bruijn or overlap graphs. This thesis introduces COGRAM (Coggins–Ramasamy Assembly Method), a genome assembly pipeline that reframes sequence reconstruction as an edge classification task on a k-mer overlap graph, replacing heuristic graph cleaning with a learned model.

COGRAM constructs a directed overlap graph from raw sequencing reads using a k-mer …

Algorithmic Trading In Idiosyncratic-Payoff Markets: A Multi-Agent System For On-Chain Prediction Contracts, Saif Aldeen A.K. Agha

CMC Senior Theses

This thesis documents the design, deployment, and forward-test evaluation of an evolutionary multi-agent algorithmic trading system on Polymarket, the largest decentralized prediction market. The system pairs a locally-hosted 72-billion-parameter language model with a gradient-boosted statistical filter and an evolutionary selection mechanism that maintains a population of approximately 500 autonomous trading agents. Each agent generates a probability estimate for an event, compares it to the prevailing market price, and trades the resulting disagreement.

The central empirical exercise estimates a panel regression of trade-level profit on the absolute disagreement between the agent's probability estimate and the market price, controlling for agent identity, …

A Specification-Driven Framework For Self-Supervised Learning In Specialized Vision Domains, Mahmut S. Gokmen

Theses and Dissertations--Computer Science

Self-supervised learning (SSL) has emerged as a principled approach to visual representation learning that derives supervisory signal directly from unlabeled data, enabling foundation models to be trained at scale without manual annotation. Deployments in medical imaging and biometric recognition have demonstrated the potential of this paradigm, yet the assumptions that make SSL effective on natural image benchmarks fail systematically in specialized domains. Generic SSL pipelines encode a tacit assumption that the most informative correspondence is spatial proximity within a single acquisition. In specialized domains this assumption breaks at the level of the data-generating process: the signal that carries domain-specific information …

Error-Driven Density Control For Compact Gaussian Splatting Under Sparse Supervision, Abdelrhman Elrawy

Theses and Dissertations (Comprehensive)

This thesis studies efficiency and stability challenges in Gaussian-splatting-based reconstruction under sparse supervision. In few-shot novel view synthesis, standard 3D Gaussian Splatting (3DGS) can overfit the limited training views and grow an unnecessarily large number of primitives due to limitations in its Adaptive Density Control (ADC) mechanism. This thesis introduces an error-driven reformulation of ADC that triggers densification using opacity gradients as a lightweight proxy for rendering error, and shows that such aggressive densification must be paired with delayed and conservative pruning to prevent destructive create--destroy cycles. When combined with depth-based geometric regularization, the resulting framework produces substantially more compact …

How Should Ai Talk About Us? Llms And Social Generics, Tiffany A. Zhu

Philosophy Faculty Publications

How should AI-generated speech balance epistemic aims, such as precision and accuracy, with ethical and social considerations? This paper examines a subtle yet consequential aspect of LLM-driven communication: the use of generic generalizations that convey information about social groups (e.g., “immigrants work low-wage jobs”). While central to human epistemic and pedagogical practices, generics are theorized to reinforce stereotypes, essentialism, and injustice. Using ChatGPT-3.5 as a case study, I uncover tendencies for AI chatbots to inconsistently hedge and refuse generics, including those that reflect well-documented social structural patterns, such as “women are more likely to get attacked while walking alone at …

Distilling The Complexity Of Agent-Based Simulations Into Textual Explanations Via Large Language Models, Noé Y. Flandre, Philippe J. Giabbanelli

VMASC Publications

Communicating the design and results of agent-based models (ABMs) to subject matter experts is challenging, which hinders participation and limits trust in simulation-based decision support. Large language models (LLMs) can communicate ABMs as textual summaries, thus complementing traditional disclosure through statistical and visualization techniques. While prior work translated the structure of conceptual models into narratives via LLMs, our extension covers the dynamics of simulation models via an automated simulation-to-text method that extracts contextual information from NetLogo ABMs, performs repeated simulations, and generates narrative descriptions (including the model’s purpose, parameters, and simulation dynamics) using mutimodal LLMs. Furthermore, four summarization algorithms spanning …

Machine Learning Classification Of Prostate Cancer Genomic Sequences Using K-Mer And Sequence-Derived Features, Kuldeep Rawat, Hirendra Nath Banerjee, Jamie Noble, Saa Naudia Deloatch, Satyendra Banerjee, Sachin Shetty, Soumya Banerjee

VMASC Publications

Prostate cancer disproportionately impacts African American men, who experience significantly higher mortality rates and earlier disease onset than other populations. Current diagnostic approaches, including prostate-specific antigen testing and biopsy, lack sufficient specificity and sensitivity, underscoring the need for accurate, molecular-level classification tools. This paper presents a machine learning framework for binary classification of genomic DNA sequences as cancerous or healthy. A dataset of 1684 FASTA-formatted sequences obtained from the National Library of Medicine - GenBank was analyzed, with 1662 sequences retained after quality control filtering. Feature engineering yielded 67 attributes, including GC content, Shannon entropy, sequence length, and trinucleotide k-mer …

Beyond Full Fine-Tuning: The New Playbook For Adapting Deep Neural Networks, Cristian S. Mcgee

Honors Undergraduate Theses

Fine-tuning is the process of teaching and specializing a pre-trained neural network on a downstream task. Fine-tuning is a rapidly growing topic in artificial intelligence domains; however, many fine-tuning endeavors are highly specialized without a coherent framework connecting them. This work presents a unified perspective on fine-tuning methods and performance metrics. Our perspective organizes the methods in terms of how they are applied to fine-tuning. This framework showcases methods that (i) update effective subspaces of the pre-trained model, (ii) change the adaptation optimization procedure, and (iii) alter the representations of the embedded input. Additionally, we present unconventional metrics such as …

An Empirical Framework For Evaluating Semantic Preservation Using Hugging Face, Nan Jia, Anita Raja, Raffi Khatchadourian

Publications and Research

As machine learning (ML) becomes an integral part of high-autonomy systems, it is critical to ensure the trustworthiness of learning-enabled software systems (LESS). Yet, the nondeterministic and run-time-defined semantics of ML complicate traditional software refactoring. We define semantic preservation in LESS as the property that optimizations of intelligent components do not alter the system's overall functional behavior. This paper introduces an empirical framework to evaluate semantic preservation in LESS by mining model evolution data from HuggingFace. We extract commit histories, $\textit{Model Cards}$, and performance metrics from a large number of models. To establish baselines, we conducted case studies in three …

Geovig And Purevig: Geometry-Aware Architectures For Efficient Computer Vision, Omar Ismail

Theses and Dissertations (Comprehensive)

Deploying deep learning models for medical image analysis on mobile devices requires a balance between inference latency, memory footprint, and delineating anatomical boundaries with high accuracy. While Convolutional Neural Networks (CNNs) and mobile Vision Transformers (ViTs) offer efficiency, they often struggle to model the irregular, non-local geometric structures inherent in biological tissues without incurring prohibitive computational costs. In this thesis, we introduce GeoViG (Geometric Vision Graph), an architecture that bridges the gap between efficient grid-based processing and explicit Geometric Deep Learning. GeoViG introduces a novel transition from high-resolution pixel grids to low-resolution dynamic graphs via a SpreadEdgePool operator, a geometry-aware …

Enlem: Ensemble Learning-Based Model To Detect Phishing Websites, Most Nilufa Yeasmin, Md Abu Rumman Refat, Bikash Chandra Singh, Zulfikar Alom, Zeyar Aung, Mohammad Azim

School of Cybersecurity Faculty Publications

Phishing involves manipulating individuals into revealing private data, e.g., user IDs, bank details, and passwords. The observed surge in fraud is related to increased deception, impersonation, and advanced online attacks. Thus, effective phishing detection methods are required to mitigate escalating global phishing threats. Existing methods (e.g., heuristics-based, signature-based, and visual similarity-based methods) attempt to detect phishing sites, and machine learning (ML) and deep learning (DL) methods are effective in the cybersecurity context in terms of learning from data, offering insights, and forecasting. However, independent ML algorithms are limited when handling complex data, and DL techniques surpass traditional ML methods in …

Explainable Physics-Based Constraints On Reinforcement Learning For Accelerator Optimization, Jonathan Colen, Malachi Schram, Kishansingh Rajput, Armen Kasparian

Data Science Faculty Publications

We present a reinforcement learning (RL) framework for optimizing particle accelerator experiments that builds explainable physics-based constraints on agent behavior. The goal is to increase transparency and trust by letting users verify that the agent’s decision-making process incorporates suitable physics. Our algorithm uses a learnable surrogate function for physical observables, such as energy, and uses them to fine-tune how actions are chosen. This surrogate can be represented by a neural network or by an interpretable sparse dictionary model. We test our algorithm on a range of particle accelerator optimization environments designed to emulate the Continuous Electron Beam Accelerator Facility at …

Untrained Position-Encoded Multilayer Perceptron Network For Structured Illumination Microscopy Reconstruction, Sahil Sharma, Leonidas Zimianitis, Krishnendu Samanta, Balpreet Singh Ahluwalia, Joby Joseph, Dushan N. Wadduwage

Computer Science Faculty Publications

Structured Illumination Microscopy (SIM) enables super-resolution imaging by encoding high-frequency spatial information through patterned light. While traditional Fourier-based reconstruction methods are prone to artifacts under suboptimal conditions, recent deep learning approaches often require large training datasets and lack adaptability across different imaging setups. In this work, we present Position Encoded Multi-Layer Perceptron (PEM) network that leverages implicit neural representations (INRs) and SIM forward-model-driven modeling to reconstruct super-resolved images without any training data. PEM-SIM represents each spatial coordinate as a combination of sinusoidal functions across multiple frequencies, enabling rich encoding of fine spatial detail. A forward model grounded in SIM image …

To Print A Remake: An Analysis Of Hollywood Remakes And Their Cultural Value, Connor F. Seaton

Honors Undergraduate Theses

In recent years, audiences and movie critics have expressed concern that Hollywood’s growing reliance on remakes, sequels, franchises, and similar adaptations has led to a broader worry that originality is fading from modern cinema and that the industry is instead focused on using adaptations to maximize profits. Although adaptation is often seen as a commercially driven framework for reproducing existing intellectual property in a new media format, this thesis argues that it should be recognized as an autonomous cultural category with its own artistic, historical, and social significance and merit. By analyzing adaptation scholarship and reviewing its complex historical development, …

Game-Based Learning For Asynchronous Ai Literacy Course: Approach To Improve Students' Cognitive, Behavioural, Affective, And Ethical Learning Of Ai, Jinhee Kim, Guang Yang, Wing Sha Chan, Xi Lin, Yukyeong Song

STEMPS Faculty Publications

Educators in higher education face persistent challenges in scaling AI literacy across disciplines and helping novice learners understand abstract AI concepts. Although research on game-based learning (GBL) reports mixed outcomes, few studies have examined its large-scale use in mandatory, asynchronous AI literacy courses for diverse undergraduate populations. Addressing this gap, this study investigates a scalable GBL-based AI literacy course delivered to 4898 first-year undergraduates across disciplines. Using a mixed-methods design with 311 valid pre- and post-survey responses and 20 interviews, the study evaluates students' cognitive, behavioural, affective, and ethical learning of AI. Quantitative results show significant improvements in overall AI …

Quantum Machine Learning Models: Principles, Frameworks, And Computational Challenges, K. A. Jayabalaji, S. Venkata Anand, Dineshkumar Rajendran, Prasanta Chatterjee Biswas, Sardor Omonov, Rubaid Ashfaq

Computer Science Faculty Publications

Quantum machine learning (QML) has become an optimistic avenue of harnessing quantum computation in data-driven modeling, especially of issues with high dimensionality and complicated correlations. Current methods are generally based on fixed or over-parameterized quantum circuits, and hence restricted to scalability as well as unproductive optimization in real-world hardware. This chapter introduces a hybrid quantum-classical learning system that is adaptive and provides principled quantum data encoding, architecture-conscious variational circuit design and resource-optimal optimization. The technique is based on the concepts of quantum architecture search and subspace-preserving transformations to trade expressiveness with trainability, and discretize the quantum model into a classical …