Digital Commons Network^™

Johnson–Lindenstrauss Projections As Architectural Components In Deep Neural Networks, Charles Q. Ray

Master's Theses

Modern deep neural networks are heavily overparameterized, with learned representations occupying lower-dimensional subspaces than their ambient dimensions suggest. This thesis investigates whether fixed random Johnson-Lindenstrauss (JL) projections can exploit this redundancy by functioning as architectural bottlenecks that reduce model size and training cost while preserving task performance.

We systematically evaluate JL-based compression across multiple architectures (convolutional networks, multilayer perceptrons, vision transformers, autoregressive language models) and tasks (reconstruction, classification, language modeling). Experiments reveal that compression tolerance depends critically on placement strategy and compression ratio.

Linear probe analysis quantifies intrinsic dimensionality at different network depths, revealing that CIFAR-10 final representations require only …

Using Ensemble Disagreement To Stabilize Conformal Prediction Under Distribution Shift, Patrick D. Murphy

Master's Theses

Semantic segmentation of eelgrass from drone imagery is crucial for coastal habitat monitoring, restoration, and management, as these habitats continue to see rapid changes due to climate change and human influence. However, the reliability of generalizing a deployed classification model relies on both high-accuracy segmentation as well as robust uncertainty quantification that holds up when conditions change over years or locations. Conformal prediction (CP) is a method that converts a classifier's output into prediction sets with a guaranteed average coverage level for in-distribution data. However, the “vanilla” conformal score can often under-cover in hard or out-of-distribution (OOD) regions under drift. …

Mechanistic Framework Of Wood Decomposition In Post-Storm Vegetative Debris Piles Using Continuous Quality Theory, Thomas T. Lee

Master's Theses

Storm events generate significant quantities of vegetative debris composed of wood, branches, twigs, and other organic materials. Efficient post-storm resource recovery and carbon accounting depend on reliable predictions of decomposition within these large, heterogeneous debris piles. While small-scale wood decay is well studied, decomposition dynamics within multi-layer debris piles remain poorly characterized, limiting municipalities' ability to estimate material longevity and resource recovery potential.

To address this gap, this study develops the WOOD_DK (Wood Organic Oxidation and DegraDation Kinetics) model, a preliminary, exploratory one-dimensional multilayer decomposition framework simulating coupled temperature, moisture, oxygen, and carbon dynamics in MATLAB. Microbial decomposition follows Continuous …

Neurosymbolic Counterpoint Generation, Paul D. Jarski

Master's Theses

Recent advancements in generative artificial intelligence have revolutionized music generation, yet research has predominantly focused on raw audio synthesis over music in symbolic form, i.e. a score. This thesis presents the first neurosymbolic model designed to generate imitative Renaissance counterpoint in symbolic (MIDI) format. By leveraging an autoregressive Transformer architecture, this research explores the capacity of deep learning models to manage independent voices and strict stylistic constraints.

We compare multiple data representation strategies with distinct tokenization methods. The proposed model incorporates a symbolic component that enforces fundamental contrapuntal rules. Additionally, this thesis contributes a preprocessed dataset of Renaissance polyphony, in …

A Simulation Framework For Edge Lane Roads And Bicycle Shared Roadways, Nick Sauciur

Master's Theses

Edge Lane Roads, also referred to as Advisory Bike Lanes, are a shared roadway treatment that designates space for cyclists while allowing flexible use of the roadway by motorists. They are commonly applied on lower speed corridors and can be used both where roadway width is limited and where agencies seek to manage vehicle speeds, clarify cyclist positioning, or improve shared roadway operations. The configuration provides visually designated space for cyclists while allowing motorists to encroach into that space when yielding to opposing traffic. Although interest in these facilities is growing in North America, quantitative tools for evaluating their operational …

Evaluating Carbon Risk And Benefit Under Improved Forest Management Prescriptions In A California Mixed Conifer Stand, Jonathan A. Garcia

Master's Theses

Improved forest management (IFM) prioritizes carbon accumulation through practices such as extended rotations and retention harvesting. These management strategies, increase fuel loads and elevate the potential fire intensity, may offset the benefits of carbon sequestration. Additionally, IFMs use spatially complex silvicultural treatments that may introduce prediction errors into the simulation processes used for long-term projections, and carbon accounting.

We evaluated the extent to which calibration in the Forest Vegetation Simulator (FVS) reduces error in aboveground carbon stock predictions and tested whether modeling approaches and calibration influence the magnitude and direction of basal area increment (BAI) prediction error using generalized mixed-effects …

Time-Dependent Amplification Of Growth Rates In A Plankton-Oxygen Model, Rapha Coutin

Master's Theses

Near-bottom hypoxia occurs when dissolved oxygen levels drop to a level that is harmful to marine biology, creating biological dead zones along the ocean floor. Recent years have seen a dramatic increase in the percentage of coastal, near-bottom, hypoxic water, with the average in 2021 nearly double that of the average from 2009 to 2018 and about twenty-eight times the average from 1950 to 1980. Recent literature has linked this increase in oceanic hypoxia to the increase in upwelling-favorable winds caused by climate change. Upwelling brings low-oxygen, nutrient-rich water up to the surface, leading to plankton blooms and mass consumption …

Video Generation Techniques For Novel View Synthesis With Flow-Matching Transformers, Xiuyuan Qiu

Master's Theses

Novel view synthesis (NVS) aims to generate images of a scene from unseen camera viewpoints. Recent work, such as Stable Virtual Camera, shows that large-scale image diffusion models like Stable Diffusion can be adapted for pose-conditioned view synthesis by incorporating video-generation techniques with camera conditioning. In this thesis, we introduce MVFlow, a new NVS model that extends this approach to a different image generation architecture: a flow-matching diffusion transformer, specifically FLUX.1, which has demonstrated strong performance in image synthesis. We evaluate MVFlow under varying input view counts and pose distance settings. Our results show that this architectural transfer is feasible; …

Design And Development Of A 12-Degree-Of-Freedom Quadruped Robot, Kai De La Cruz

Master's Theses

This thesis presents the design, analysis, and experimental validation of a 12-degree-of-freedom (DOF) quadruped robot developed as a research platform for legged locomotion and control. The system builds upon the Cal Poly Legged Robotics Group’s prior 8-DOF quadruped, addressing key limitations in manipulability, load distribution, and mechanical robustness by introducing a 3-DOF leg architecture.

The mechanical design emphasizes lightweight construction, structural integrity, and modularity to support future research extensions. Static and dynamic loading models were developed to inform component sizing and material selection. A carbon-fiber chassis and redesigned leg assemblies were fabricated and validated through finite element analysis and physical …

The Application, Construction, And Validation Of Hidden Markov Model Profiles For Carbonic Anhydrase Enzymes, Samuel F. Kaplan

Master's Theses

Carbonic anhydrases (CAs) catalyze the reversible hydration of CO2 and have evolved independently at least eight times, resulting in structurally distinct enzyme families (α, β, γ, δ, ζ, η, θ, ι). Traditional sequence alignment methods struggle to classify these convergently evolved proteins because their sequential similarity does not reliably indicate functional or evolutionary relationships. Many CA sequences in public databases are annotated generically without family assignments, and prior computational approaches have focused predominantly on the three well characterized families (α, β, γ), leaving the five recently discovered classes without robust classification tools. Family level assignment is often a prerequisite for …

Sea Level Trends Near Critical Levees In The Sacramento-San Joaquin Delta, Reyna Sanchez Gomez

Master's Theses

The Sacramento-San Joaquin Delta (Delta) levee system faces threats from rising sea levels, land subsidence, and deteriorating levees, all of which pose risks to California’s primary surface water supply. The purpose of this thesis is to assess relative sea level rise (RSLR) and estimate the spatially varying factors that influence water levels in the Delta. This thesis extends the work of Baranes et al. (2023) to the period 1975-2022 by appending recently acquired water-level records (1975-1982) to modern datasets (1982-2022) from 10 water-level stations. These water level datasets were digitized, quality-assured, and referenced to a common geodetic datum, as described …

Covering Spaces, Mapping Class Groups And Heegaard Splittings, Tri Tran

Master's Theses

One studies the symmetries of an algebraic object by looking at its automorphism group. For example, the automorphism group of a regular n-gon is generated by rotations and reflections, and the automorphism group of {1,...,n} is generated by transpositions. We study the group of symmetries of a closed, oriented 2- and 3-dimensional manifold M, called the mapping class group of M. In the 2-dimensional case, it is known explicitly what the generators of the mapping class group are. We present a proof of this fact following Farb and Margalit's exposition. The question about the mapping …

Exploring Non-Conventional Substrates: The Role Of Hibiscus And Whey In Kombucha Fermentation, Benjamin J. Lee

Master's Theses

This study evaluated the impact of hibiscus and whey as alternative fermentation substrates on the physicochemical, phytochemical, and microbial properties of kombucha. A 2 × 2 full factorial design was used to compare four treatments: black tea + water (BW), black tea + whey (BH), hibiscus + water (HW), and hibiscus + whey (HH). The fermentation solvents (water or whey) were heated to 88 °C, infused with 0.6% (w/v) black tea or hibiscus for 10 minutes, sweetened with 10% (w/v) sucrose, cooled to 22 °C, and inoculated with 20% (v/v) kombucha starter culture. Fermentations were carried out at 22 °C …

Measurement And Control Of The Oxidation-Reduction Potential And Its Chemical And Sensory Consequences In Wines, William Jordan Wright

Master's Theses

The oxidation-reduction potential (ORP) is a measurement thought to reflect the overall ability of a solution to oxidize or reduce substances within that solution. In wine, there is no absolute agreement as to which substances control or dominate the ORP; however, oxygen, ethanol, iron-tartrate and copper-tartrate complexes, and glutathione are suspected to play key roles, while sulfur dioxide and phenolics may act as buffers. Anecdotally, some varieties of Vitis vinifera are prone to oxidation while others are prone to reduction, though the chemical basis of this remains unclear. The following works aim to further elucidate the influence of specific chemical …

Designing For Trust In Chat-Based Question Answering Systems: An Exchange-Based Retrieval Approach, Nathan Mccutchen

Master's Theses

Community chat platforms such as Discord and Slack support spontaneous, collaborative communication but make it difficult to retrieve previously discussed information. As conversations accumulate, valuable exchanges become buried, leading to repeated questions and sustained burden on experienced community members.

This work contributes a set of design requirements for question-answering systems operating over unstructured chat data, a Discord bot prototype implementing those requirements named Echo, and an empirical evaluation of how such a system affects user trust. Rather than encoding discrete question-answer pairs or generating synthetic responses with a language model, Echo indexes conversation topics for semantic retrieval and presents results …

Associations Between Nutrition-Related Chronic Disease And Food Insecurity: Data From The California State University, Elizabeth Ochoa

Master's Theses

Food insecurity is a growing public health concern consistently associated with elevated rates of nutrition-related chronic disease (NRCD) across the general population. Despite young adulthood being a critical window for dietary pattern formation, this relationship remains minimally explored among college students. This thesis examined associations between food insecurity and NRCD diagnosis, including hypertension, diabetes, hyperlipidemia, and cardiovascular disease, among undergraduate students across the California State University (CSU) system. Cross-sectional data were drawn from the Spring 2021 CSU National College Health Assessment, administered across all 23 campuses. After exclusions, the analytical sample comprised 13,762 students. Bivariate analyses assessed relationships between key …

A Ship Between Worlds: Illicit Trade, Empire, And Enslavement In The Voyage Of The Nuestra Senora De Las Nieves, 1819-1820, Carly Eastham

Master's Theses

This study examines the illicit Spanish transatlantic slave trade to Cuba during the post-abolition era through a microhistorical analysis of the 1819 voyage of the schooner Nuestra Señora de las Nieves. Despite international efforts to suppress the trade following abolition treaties, this voyage illustrates how Spanish traders adapted their practices to continue trafficking enslaved Africans, fueling Cuba’s booming sugar economy amid shifting geopolitical and economic pressures. Drawing on primary documents seized by the British navy and analyzed alongside relevant secondary scholarship, this research situates the Cuban trade within the broader Atlantic world’s complex cultural, legal, and economic transformations.

The case …

A Scalable Iteration Of The Horizon Simulation Framework Using Multithreading Techniques, Jason E. Beals

Master's Theses

The Horizon Simulation Framework (HSF) occupies a unique space in the modern aerospace modeling landscape, enabling flexible, modular modeling of mission-level agent behavior through an object-oriented, hierarchical design. HSF's hallmark breadth-first search scheduling algorithm explores a "multiverse" of possible mission execution pathways, enabling exhaustive evaluation of schedule combinations against user-defined heuristics.

As aerospace systems become increasingly complex, HSF faces critical challenges in establishing verifiable, deterministic behavior. The framework's core scheduling algorithm had not undergone systematic validation, leaving questions about temporal consistency, state management correctness, and reproducibility across different program executions. Furthermore, the exponential growth of schedule combinations creates computational bottlenecks …

Empirical Comparisons Of Partial Dimension Reduction Algorithms In High-Dimensional Regression, Nathan Greenfield

Master's Theses

In high-dimensional regression problems, dimension reduction methods are often used to address the challenges of multicollinearity and estimation instability. Partial dimension reduction extends these ideas by applying dimension reduction to a subset of the predictors, while the remaining predictors are modeled without compression. This approach is particularly useful when it is important to retain variability and interpretability in certain predictors.

This thesis investigates the empirical performance of partial dimension reduction algorithms and introduces a novel algorithm, Iterative Partial Residual (IPR). Two algorithms are considered: a baseline algorithm, Marginal Residual (MR), and the proposed IPR method. Their predictive performance is evaluated …

Opioid Use Knowledge And Attitudes Among U.S. Adults, Jadyn E. Knox

Master's Theses

The opioid epidemic remains a public health crisis, with overdose rates continuing to rise. Associations between individuals opioid use and knowledge of and attitudes toward opioid substances and overdoses have been investigated to inform educational interventions. However, many studies utilize college student or adolescent samples, and findings remain inconclusive. An assessment of opioid knowledge and attitudes across broader adult populations is lacking. Concurrently, replication literature in the social sciences is limited, therefore, the present study replicated and expanded upon Williams et al. (2013) and Vyas et al. (2023) studies. Reliability and validity of the Opioid Knowledge scale, Opioid Overdose Knowledge …

A Literature Review On Ethics In Medical Diagnostic Ai & Analysis Therein, Connor J. Turetzky

Master's Theses

Artificial Intelligence presents a very promising future in medicine. Being able to diagnose and recommend treatments quickly is vital in ensuring positive patient outcomes. However, the new technology is not without risk. In this narrative literature review, the risks of AI in terms of bias, ethics, and environmental impact will be explored through existing research. This paper will focus on research published between 2019 and 2026, highlighting the major ethical and systematic problems currently facing diagnostic AI. Historical bias in medical data has led to AI that share those biases, and humans inherit that bias creating a potential negative feedback …

Defining Limited English Proficiency: Implications For Assessment And Special Education Placement Decisions, Christine Vernon

Master's Theses

Problem

The problem addressed was the lack of a clear, consistent definition of Limited English Proficiency, despite its influence on funding, programming, assessment, and student placement. The purpose of the study was to clarify how LEP is constructed and used in educational literature and to explore its impact on assessment practices. This study examines the definition of Limited English Proficiency and its implications for educational assessment and special education placement decisions.

Methodology

An integrative literature review, guided by meta-narrative methodology and structural coding, was conducted using 119 peer-reviewed sources from APA PsycInfo, ERIC, and JSTOR (2001–2023).

Findings

Findings revealed that …

Assessing The Ecological Drivers Of Wildfire Hazard Mitigation Of Fuels In The California Home Ignition Zone, Christopher David Benterou

Master's Theses

Consecutive years of devastating home losses to wildfire indicate the need for home fuel reduction and construction improvements. Existing research correlates house survival with modifications to the home ignition zone (HIZ) and is required in high fire hazard areas of California (CA). Mitigations, designed to disrupt continuous fuel patterns, (e.g., separating surface and canopy fuels), can also introduce ecological changes that can alter overall code compliance. Previous compliance studies used respondent and agency-generated data, which may lack accuracy. Assessing additional house and ecological features (e.g., canopy cover and construction) would improve prediction of compliance, but these relationships are poorly studied. …

Exploring Math Word Problem Generation With Llms, Trung Hieu Vuong

Master's Theses

Math Word Problem (MWP) is an important building block for learning math. This type of problem is particularly useful for younger audiences because solving it involves two simultaneous skill sets: reading comprehension and mathematical reasoning. With publicly available large language models (LLMs), generating additional MWPs is readily achievable. While researchers have started using LLMs as MWP facilitators, there still exists a gap in studies about the diversity of MWPs generated by unmodified, publicly accessible LLMs. For that reason, our study focused on two goals: (1) to evaluate the diversity of MWPs generated by publicly available LLMs when provided with examples …

Developing A Vietnamese Text Summarization Large Language Model On Limited Hardware, Tin Pho

Master's Theses

Text summarization models have achieved significant growth during the last few years because of major Large Language Model (LLM) technological advancements. The application of LLMs are widely used in news distribution (TL;DR news), translation tools (DeepL Translate), or virtual assistants (ChatGPT, DeepSeek, Claude, etc.). However, the progress has not yet reached all languages equally. The Vietnamese language is used by more than 90 million people, but the language is not as highly developed for LLM as it has many homophones, five different tones that affect meaning of words, and irregular grammar compared to other languages (e.g. English, Spanish, etc.). Also, …

Estimating Prey Capture Attempts And Activity Costs In Foraging Emperor Penguins (Aptenodytes Forsteri) Using Integrated Biologging Data, Taylor Rose Azizeh

Master's Theses

To forage efficiently in patchy and unpredictable prey landscapes, marine predators must optimize energy intake while minimizing costs. Central place foragers like emperor penguins face greater energetic constraints during demanding periods of high investment like chick-rearing. I used high-resolution tri-axial acceleration, depth, and GPS data from 25 birds from two feeding seasons (2019 and 2022) to quantify prey capture attempts and a proxy for energy expenditure (overall dynamic body acceleration; ODBA) across dive types (epipelagic, mesopelagic, benthic), phases (descent, bottom, ascent), and over time at-sea. While prey capture attempts per minute did not differ when looking across dive types alone, …

Automatic Guitar Transcription Of Polyphonic Music, Ritwik Patil

Master's Theses

Transcribing guitar music automatically is a complex task due to polyphonic overlap, tuning variations, and diverse playing techniques. Current transcription systems focus on identifying note pitches and timing while ignoring performance techniques that describe how the notes are played, treating guitar recordings as generic polyphonic audio and producing MIDI-like outputs that lose important information about articulation and style. To address these challenges, we propose an end-to-end transformer model for automatic guitar transcription. The system uses a T5-based encoder-decoder architecture that processes the Constant-Q Transform (CQT) of stereo audio input. The stereo representation helps separate individual guitar parts within a mix …

Mobile Markets And Farmers’ Markets: Evaluating Food Accessibility For Underserved Communities In The Bay Area, Rajeshwari Rajesh

Master's Theses

Studies indicate including farmers’ markets, mobile markets, and other local food systems improve fresh food access for low-income communities, especially when paired with food assistance programs that offer financial incentives. This study aims to evaluate mobile markets and farmers’ markets in the Bay Area using the five dimensions of food access: accessibility, availability, acceptability, affordability, and accommodation. The study is based on primary data collected from 173 in-person surveys at six mobile and eight farmers’ markets in the Santa Clara and Contra Costa counties. Key survey participants were market shoppers participating in food assistance programs. A mixed research methodology included …

Latent Action Trajectory Optimization, Rahul Milind Kandekar

Master's Theses

Learning from demonstrations offers a path to bypass the sample inefficiency of reinforcement learning, but obtaining action-labeled expert demonstrations remains expensive and often impractical. Learning from Observations (LFO) addresses this by learning policies from observation-only demonstrations. Recent LFO work relies heavily on behavior cloning: VPT and LAPO use observation-only data combined with limited action labels to train BC policies, while AIME offers an alternative policy inference approach but requires the majority of its training data to have action labels. Through systematic experiments in the Lunar Lander environment, we investigate whether latent action methods can function when state and action dimensionalities …

Game Genre, Gender, And Sense Of Community In Online Gaming, Patrick Brown

Master's Theses

Sense of community has been shown to improve quality of life in numerous ways. Previous research suggests sense of community can be felt even within groups that never meet in person. The current study explored the relation between game genre, gender status, and sense of community in online gaming groups. Additionally, discrimination was included as an additional outcome due to its close link with sense of community. A total of 189 college students (Mage = 20 years old) of diverse gender and ethnicity completed an online survey consisting of the Sense of Community Scale, Everyday Discrimination Scale, demographic questions, and …