Physical Sciences and Mathematics Commons^™

Cost-Aware Predictive Routing For Vision-Language Models, Sahil Sait Naveed

Master's Projects

Vision-language models (VLMs) are increasingly used for multimodal tasks, but their inference costs vary widely across model tiers. This work presents a predictive routing framework that assigns each query to the most cost-effective VLM by using multimodal embeddings, clustering, and per-cluster model error estimates. Evaluated on six commercial VLMs, the selected router reduces average cost by about 49 percent relative to the quality-first setting on both validation and test. This comes with accuracy losses of 0.4 percentage points on validation and 0.9 on held-out test data. It also outperforms K-NN and ZeroRouter routing baselines while providing lowercost operating points than …

Adaptive Real-Time Fraud Detection Using Online Learning And Explicit Concept-Drift Detection, Purva Govind Tugaonkar

Master's Projects

Real-time credit card fraud detection faces challenges such as extreme class imbalance, delayed feedback, and concept drift in transaction streams. This project implements and evaluates an adaptive streaming fraud detection framework based on three methodologies: (1) online learning with incremental updates, (2) explicit conceptdrift detection using statistical monitoring, and (3) separate models for immediate and delayed supervision, combined with cost-sensitive learning and anomaly detection. The system processes the credit card fraud dataset in a batched streaming fashion, uses multiple online learners and ensembles. Experiments show that online, driftaware models maintain high recall on frauds while controlling false positives under imbalanced …

Modeling Sealed Deck Construction In Collectible Card Games Using Learning-To-Rank Approach, Michael Dinh Nguyen

Master's Projects

Artificial intelligence has demonstrated strong performance in complex decision-making domains such as chess and Go, motivating research into its application for games with even richer rules and combinatorial complexity. In collectible card games like Magic: The Gathering, deck construction from a constrained card pool is a critical and challenging task that requires evaluating card strength, synergy, and resource balance. This project explores whether machine learning, specifically learning-to-rank (LTR), can effectively model these human decision processes to construct competitive decks in a sealed format. The results found here can also be applicable to other areas of note, such as sports drafting …

Multi-Agent Path Planning And Optimization Using Q-Learning, Chirag Rudresh

Master's Projects

Robot navigation in a multi-agent setting requires a balance between safety and efficiency, especially in dense environments. In these two-dimensional spaces, the scope for geometric errors is much less and could lead to collisions or immobility. This project proposes to address the navigation task using a two-phase path planning pipeline that combines reinforcement learning and convex optimization in a scalable and robust manner. The first phase consists of generating diverse collision-free paths using a Q-learning agent that is trained on a visibility graph representation of the environment. The discretization of the environment using waypoint-based graphs allows the agent to train …

Challenges And Applications Of Fine-Grained Temporal Action Understanding: Modeling Temporal Granularity And Data Efficiency, Halil I. Helvaci

Theses and Dissertations--Electrical and Computer Engineering

Fine-grained Temporal Action Segmentation (TAS) has become a cornerstone of video understanding, offering dense frame-level predictions essential for clinical assessment, surgical skill evaluation, and human-computer interaction. While TAS methods have delivered strong results on coarse-grained benchmarks, two fundamental challenges persist: (1) global attention mechanisms dilute boundary information critical for subsecond precision, a phenomenon we term the temporal granularity bottleneck, and (2) dense frame-level annotation remains prohibitively expensive, with most datasets requiring exhaustive labeling of lengthy untrimmed videos. These challenges are particularly pronounced in medical domains, where sub-second primitives define clinical outcomes while expert annotation remains scarce. In this dissertation, we …

Polysemicolon; Novice Programmers And Java Keywords, Briana C. Bettin

Michigan Tech Publications

Most industrial programming languages leverage the English language for reserved keywords – words which a program compiler recognizes as specific execution commands. The divide between expert and novice programmers showcases an intriguing middle-ground by which the polysemy of many keywords becomes revealed. This essay explores a sampling of the multitude of keyword interpretations that a novice programmer may derive from the Java language’s syntactic style and keywords specifically, and how the polysemy of both “English” and “code” meanings to these terms affects the novice-expert programmer transition.The transition from novice to expert, and the mapping of the career of metaphor to …

Analysis Of The Thermodynamic Variables Associated With Hawaiian Extreme Precipitation Using The Fraction Of Attributable Risk (Far), Matthew J. Sinnenberg

Electronic Theses & Dissertations (2024 - present)

The Hawaiian Islands experience extreme maxima in precipitation that vary spatially, by season, and in response to different teleconnection patterns. Due to the islands’ unique geography, they are not well-resolved by global climate models, necessitating a focused assessment to identify  climatological trends. In this study, a probability-based analysis was conducted on regionally averaged thermodynamic variables relevant to the formation of extreme events. The primary objective was to evaluate whether changes in the occurrence of extreme conditions have occurred and to understand how the El Niño-Southern Oscillation (ENSO) influences these occurrence rates.

Hourly data from ERA5 spanning 1980-2019 were resampled to …

Nanostructured Cathode Catalysts For Aem Electrolysis: From Catalyst Design To Degradation And Hydrogen Dynamics, Yamini Kumaran

Electronic Theses & Dissertations (2024 - present)

Anion exchange membrane water electrolysis (AEMWE) presents a promising pathway toward cost-effective and sustainable hydrogen production by integrating the chemical robustness of alkaline systems with the compact, zero-gap design of proton exchange membrane electrolyzers. However, the widespread implementation of AEMWE is limited by the availability of highly active and durable platinum-group-metal (PGM)-free catalysts and by an incomplete understanding of their degradation behavior under realistic operating conditions.

This dissertation focuses on the development, characterization, and mechanistic investigation of nanostructured MoNi₄–MoO₂-based electrodes for efficient and stable hydrogen generation under alkaline and membrane-integrated environments. MoNi₄–MoO₂ nanorods …

Advancing Periodic Time Series Analysis: Application, Bias Assessment, And Optimal Window Selection In The Variable Bandpass Periodic Block Bootstrap Method, Yanan Sun

Electronic Theses & Dissertations (2024 - present)

Time series analysis is essential for understanding long-term patterns, periodic behavior, and underlying correlations in complex datasets. The periodically correlated (PC) time series is a type of time series where the correlation structure repeats over fixed intervals. The Variable Bandpass Periodic Block Bootstrap (VBPBB) has recently been proposed as a resampling method that preserves PC structures through the use of periodogram, bandpass filters, and block bootstrap resampling. Although promising, VBPBB remains underutilized, and its limitations have not been fully examined. This dissertation advances both the application and methodological development of the VBPBB.

The first project applies the VBPBB to a …

Graph-Based And Graph-Transformer Representation Learning For Healthcare Data, Rui Wang

Electronic Theses & Dissertations (2024 - present)

Healthcare data exhibit complex structures, including heterogeneous clinical entities, sparse observations, and longitudinal patient trajectories. Effectively modeling such data remains a fundamental challenge in computational healthcare research. Traditional machine learning approaches often rely on flat feature representations that fail to capture relationships among clinical events, limiting their ability to model complex healthcare processes. These challenges motivate structured learning frameworks that capture both relational structure and temporal dynamics in healthcare data. This dissertation develops a series of graph-based representation learning approaches, extended through graph-transformer architectures for modeling complex healthcare data. Such data can be represented as graphs, where nodes correspond to …

The Global Orbit ∞-Category And Applications To Assembly Maps, Zoë Pope

Electronic Theses & Dissertations (2024 - present)

We reformulate the foundations of assembly maps in the context of the global orbit ∞-category of all discrete groups. We first show that the ∞-categorical slice of the global orbit ∞-category over any fixed group G is equivalent to the orbit 1-category of G, and we also frame the subgroup 1-category of G in this global context. We then use the afore-mentioned equivalence to redefine assembly maps as counits of the adjunction between left Kan extension and restriction, and give purely ∞-categorical and conceptual proofs of known results, such as the Transitivity Principle. Additionally, we give equivalent formulations for what …

Developments And Applications Of Transport Of Intensity Equation Imaging And Optical Diffraction Tomography In Bioimaging, Shane D. Carney

Electronic Theses & Dissertations (2024 - present)

Non-invasive imaging schemes are highly sought after methodologies in biology and metrology, where maintaining the physical state of a sample is important. Quantitative phase imaging (QPI) techniques non-invasively reconstruct the relative phase delay imparted onto an optical field due to transmission through a sample. Phase delay is proportional to the thickness and refractive index of the sample, but is lost in the intensity measurement process. Digital holographic microscopy (DHM) and transport of intensity equation (TIE) phase imaging are two high-resolution QPI techniques. DHM is a well-researched, accurate, and widely implemented technique, but can be difficult to implement in labs without …

Retrieval And Characterization Of Aerosol Optical And Microphysical Properties Following Pyrocumulonimbus Events From Sageiii/Iss, Vishal G. Bagadia

Electronic Theses & Dissertations (2024 - present)

Stratospheric aerosols play an important role in maintaining global radiative balance with impacts to stratospheric equilibria through physical and chemical forcings.  Stratospheric aerosol size distributions, in particular aerosol size, are vitally important in determining the physical and chemical impacts of aerosols in the stratosphere. The representation of aerosol size and size distributions in most climate models contain large uncertainties due to relatively sparse sampling of stratospheric aerosols. Occultation measurements from satellite provide an unique opportunity to retrieve the aerosol size distribution using high quality aerosol extinction measurements. A retrieval algorithm is developed to characterize size distributions of stratospheric aerosols from …

Entropic Foundation Of Finance And Physics: Securities Price Dynamics And Quantum Theory, Mohammad Abedi

Electronic Theses & Dissertations (2024 - present)

In many scientific and financial contexts, we must reason and make predictions under conditions of incomplete information. This dissertation develops Entropic Dynamics (ED) as a unified framework for deriving dynamical laws directly from principles of inference. Within this approach, probability distributions represent states of knowledge, and their evolution is determined through entropy maximization subject to relevant constraints. This leads to a novel concept of entropic time and a formulation of dynamics as an inferential process. In this talk, I will present how ED provides a common foundation across multiple domains. In physics, quantum dynamics for particles and scalar fields in …

An Evaluation Of Current And Future Extreme Precipitation Events In The South American Altiplano Using Convective-Permitting Model Simulations, Karen R. Conron Chamberlain

Electronic Theses & Dissertations (2024 - present)

Extreme precipitation events in the South American Altiplano present significant hazards to communities in the Andean region, yet their future characteristics under anthropogenic climate change remain uncertain because of complex terrain and limitations of course-resolution climate models. This study evaluates the ability of South America Affinity Group Weather Research and Forecasting (SAAG-WRF) model with 4-km grid-spacing to reproduce 2000-2015 temperature, precipitation, and extreme precipitation events over the Altiplano. Examination of projected future changes is completed using the SAAG-WRF pseudo global warming (WRF PGW) simulation representing an approximately 3°C warmer global climate. The historical/control simulation (WRF CTL) was evaluated against DECADE …

When Helpfulness Becomes Harmful: Jailbreaking Llms For Malicious Code Generation, Noelle Capodieci

Electronic Theses & Dissertations (2024 - present)

Generative AI (GenAI) and Large Language Models (LLMs) have made large strides in coding task capabilities, with many software developers integrating agentic engineering into their workflow. While GenAI has largely benefited professional software engineers who can automate their work, it has also created room for those with little coding expertise to also create fully fledged programs and applications. It is commonly noted that GenAI is trained with two major goals in mind: to be as helpful as possible, and be as harmless as possible. There exist moments where helpfulness may be prioritized over harmlessness when these goals conflict. LLMs may …

The Predictability And Verification Of Precipitation From Landfalling Tropical Cyclones Using An Object-Based Framework, Melissa Piper

Electronic Theses & Dissertations (2024 - present)

Precipitation from landfalling tropical cyclones (TCs) poses a significant risk to life and property in both coastal and inland communities. Early warning systems can help mitigate the impacts of TC-induced flooding; however, this requires accurate quantitative precipitation forecasts (QPFs). Forecasting precipitation from landfalling TCs is a challenge, as the intensity, duration, and location of precipitation can be dependent on numerous storm-related and environmental factors. Previous studies have attempted to identify and correct NWP model deficiencies and reduce forecast uncertainty by applying traditional and advanced verification metrics to TC QPF. However, these studies have mostly used gridpoint verification metrics, limited samples …

Hydroclimate Variability And Itcz Migration Over The Last Millennium Inferred From A New Speleothem Record In Central Colombia Using Paleo Data Assimilation, Derick Appiah Kubi

Electronic Theses & Dissertations (2024 - present)

Hydroclimate variability over tropical South America is strongly influenced by meridional shifts in the Intertropical Convergence Zone (ITCZ), yet the spatial expression and dynamical controls of these changes during the last millennium remain poorly understood. This work investigates hydroclimate variability and ITCZ migration over the last millennium (850-1850 CE) using a new isotope-enabled data assimilation (DA) reconstruction made from a suite of speleothem δ¹⁸O records from the neotropics on both sides of the ITCZ. The reconstruction targets the March-August (MAMJJA) season in order to capture the rainfall regimes most relevant to central Colombia and northeastern Brazil. The focus on these …

Weather-Related Road Condition Detection Using Co-Developed Machine Learning Methods, Carly Sutter

Electronic Theses & Dissertations (2024 - present)

Transportation agencies make critical operational decisions during hazardous weather events, including assessment of road conditions and resource allocation. In this study, machine learning models are developed to provide additional support for the New York State Department of Transportation (NYSDOT) by automatically classifying road conditions across the state. Convolutional neural networks and random forests are trained on NYSDOT roadside camera images and weather data to predict road surface conditions. This task draws critically on a labeled dataset of 22,000 camera images containing six road surface conditions (severe snow, snow, wet, dry, poor visibility, and obstructed) generated from an iterative hand-labeling process …

The Ellipsoidal Principal Semi-Axis Geometry Of The Solution To An Ivp For A Matrix Diffusion Pde, Brandon M. Fox

Electronic Theses & Dissertations (2024 - present)

We analyze the underlying geometry of the solution to an IVP for a matrix diffusion PDE. We first derive the fundamental solution to the PDE. We then determine the unique solution to the IVP. From there, we begin analyzing its underlying geometry. We first observe that the geometry exhibits an ellipsoidal nature. Furthermore, we observe that it is described by the principal semi-axis geometry of the ellipsoids associated with the solution. This conclusion follows from applying the Principal Axis Theorem to the associated ellipsoids to establish their principal semi-axis geometry, as governed by the eigenstructure of the matrix. This thesis …

Recent Climatology And Changes In Precipitating Storm Characteristics Over Conus, Tyler Landsparger

Electronic Theses & Dissertations (2024 - present)

As anthropogenic greenhouse gas emissions continue to warm the global climate system, improving our understanding of how precipitating storms are responding in the real world has become increasingly important.  This study leverages 22 years of high-resolution hourly precipitation data from the NCEP Stage IV dataset to produce a seasonal climatology and long-term trend analysis of precipitating storm frequency, duration, area, and intensity across the Contiguous United States (CONUS) from March 2002 through February 2024.  Storms were identified and tracked using an object-tracking algorithm and their characteristics were computed for each season, from which 22-year climatology and trends were derived.  The …

Tracking A Precessing Quantum Spin System: Extracting Data From Quantum Systems Using Bayesian Analysis, John Kaminski

Electronic Theses & Dissertations (2024 - present)

The parameters that govern the evolution of a quantum system can be determined by making measurements on an ensemble of identically-prepared systems. However, there are times when we are unable to recreate the experiment and want to measure physical parameters from just a single particle. For example, we may wish to measure the mass to charge ratio of a particle based on its track through a bubble chamber. Recent research [11, 12, 16] has suggested that classical behavior, defined as phenomenon that does not violate macroscopic realism, does not arise simply as a consequence of a quantum system of many …

Generative Ai Peer Tutoring To Support Peer-Reviewed Source Identification And Evaluation, Rae Mair, Michelle Kelley, Taylar Wenzel, Andrea C. Borowczak, Yuqing Li, Mike Borowczak

Faculty Scholarship and Creative Works

This chapter explores the design and evaluation of a generative artificial intelligence peer tutor prompt to support college students in identifying and evaluating peer-reviewed sources for academic research. Grounded in literature on peer tutoring, Socratic dialogue, and AI-supported learning, the authors describe an iterative prompt engineering process designed to transform large language models (LLMs) into Socratic-style peer tutors capable of scaffolding student reasoning without completing tasks for them. Five guiding criteria for an effective peer tutor shaped development and evaluation: cognitive congruence, step-by-step guidance, avoiding giving answers, adaptability to student level, metacognitive transparency, and following assignment directions. Across multiple human-centered …

Brst Conserved Moments For Fluids, James Glimm

Department of Applied Mathematics & Statistics Faculty Publications

This note adds a missing proof for an assertion in the original paper Gauge Invariance and Repeated Time Isolated Solution Discontinuities.

The assertion is that under weak limits and within a constant energy isosurface, enstrophy is conserved up to losses due to viscous dissipation.

According to the BRST theory for quantum Yang-Mills fields, the conserved moments are exactly those with infinite vacuum expectation values. Application of the BRST theory to fluids is explored. The BRST restriction to the loop expansion (fixed space) is justified by the reqirement that a perturbative expansion can be constructed. In this case the theory predicts …

Weak Solutions Of The Navier-Stokes Equation And Their Euler Limits, James Glimm

Department of Applied Mathematics & Statistics Faculty Publications

The existence of a weak solution of the incompressible isothermal Navier-Stokes equation with given initial conditions in the Sobolev space $\mathcal{H}_{-2}$ for energy fluctuations and in $\mathcal{H}_{-3}$ for enstrophy fluctuations is assumed. The existence is uniform with respect to the Euler limit of zero viscosity $\nu$. Thus, existence of weak solutions of the Euler equation with given initial conditions is established, and these Euler solutions are the zero viscosity limit of Navier-Stokes solutions, provided the Navier-Stokes solutions exist.

The Yang-Mills Field: The Fermion Sign Problem, James Glimm

Department of Applied Mathematics & Statistics Faculty Publications

A new resolution of the Fermion sign problem is accomplished based on the theory of homology groups and thimbles. The sign problem for residual states not reached by the homology groups and thimbles is resolved by a perturbative converged Bosonic expansion. The proofs depend on an assumed principle of a maximum rate of entropy production. Two Yang-Mills gauge field theories are constructed, one based on short distance asymptotics and the other based on long distance asymptotics.

Quantum General Relativity V2, James Glimm, James Glimm

Department of Applied Mathematics & Statistics Faculty Publications

We construct a quantum general relativity theory.. The construction includes specification of a cosmological model. Based on type Ia supernova data, the model has zero dark energy and zero cosmological constant.

The analysis is at the level of theoretical physics, with some mathematical details omitted.

The Pure Yang-Mills Field: I: Fixed Time Existence, James Glimm, James Glimm

Department of Applied Mathematics & Statistics Faculty Publications

The convergence of renormalized perturbation theory to all finite orders is defined and shown to be valid for the fixed time perturbation theory of a pure Yang-Mills field.

Two pure Yang-Mills quantum gauge field theories are constructed, one based on short distance asymptotics and the other based on long distance asymptotics.

The construction depends on an assumed principle of a maximum rate of entropy production.

Non-Smooth Solutions Of The Navier-Stokes Equation And Their Means, James Glimm, Jarret Petrillo

Department of Applied Mathematics & Statistics Faculty Publications

Non-smooth (finite time blowup) Leray-Hopf solutions of the incompressible Navier-Stokes equation are constructed. The initial data for blowup is characterized by nonzero energy related turbulent fluctuations. The construction occurs in a finite periodic cube T3. The mean value of a weak solution of the Navier-Stokes equation is identified as a smooth solution of the Navier-Stokes equation.

Impact Of Land Use On Soil Properties, Clay Mineralogy, And Behavior Of Ammonium Across A Precipitation Gradient, Isabel Delamater

Theses and Dissertations--Plant and Soil Sciences

Like soil texture, clay mineralogy is considered a permanent or inherent characteristic of soils reflecting the influence of soil forming factors over long periods of time. Despite the importance of clay mineralogy in providing ecosystem services, much remains unknown about whether land use (native grassland sod versus no-tillage) can modify clay mineralogy under short periods of time. Furthermore, in no-tillage agroecosystems, lower crop yields are often ascribed to immobilization of added N fertilizer where inorganic nitrogen (such as ammonium) is converted to organic nitrogen. The possible role of clay mineralogy in retaining added ammonium has been overlooked in no-tillage systems. …