Data Science Commons^™

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, …

A Novel Framework For Dynamic Graph Representation Learning With Mamba, Ashish Pandey

Theses

Dynamic graph embedding is a key technique for modeling temporal dependencies in evolving networks. While transformer-based models perform well, their quadratic complexity limits scalability on long graph sequences. This thesis compares transformer approaches with the Mamba architecture-a linear-complexity state-space model—for temporal graph embedding.

Two frameworks are proposed: DG-Mamba and GDG-Mamba. DG-Mamba uses standard GCN-based spatial encoding, while GDG-Mamba incorporates domain-aware edge features using Graph Isomorphism Network with Edge Convolution (GraphGINE). Experiments on UCI, Reality Mining, Slashdot, Bitcoin-OTC, and SBM datasets show that Mamba-based models match or exceed transformer performance, especially on graphs with high temporal variability.

The thesis also applies …

Advancing Prediction Of Stimulant Medication Misuse Through Graph Representation Learning, Hamid Razavi

Theses

The misuse of stimulant prescription medications poses a significant and escalating public health concern in the United States, particularly among young adults. Addressing this issue requires sophisticated methodologies capable of uncovering complex patterns and relationships in data. Geometric Deep Learning, a paradigm designed to analyze data with non-Euclidean structures, has achieved remarkable success across various domains, offering a powerful framework for tackling complex graph structure data challenges.

This study leverages Graph Convolutional Networks (GCNs) to predict the likelihood of stimulant medication misuse using data from the National Survey on Drug Use and Health (NSDUH). Individuals are represented as nodes in …

Applications Of Neural Networks In Parkinson’S Disease Diagnosis, Saladin Minhaaj

Theses

Parkinson's disease (PD) is a complex and debilitating neurodegenerative disorder that affects millions of people worldwide. Early and accurate diagnosis is crucial for effective treatment and management of PD. This thesis explores the application of neural networks in PD diagnosis, leveraging their ability to learn patterns from large datasets and make accurate predictions.

Thesis provides an overview of PD, including its symptoms, diagnosis, and current challenges in diagnosis. We then delve into the fundamentals of neural networks, including supervised learning, mathematical interpretations, and parametric models. This research focuses on the development of neural network models that can accurately diagnose PD …

Holistic Correlation Measure For Enhanced Encapsulation Of Trait Heterogeneity And Discovery Of Co-Expression, Zachary Valleroy

Theses

Large-scale, high-dimensional data analyses can be computationally prohibitive due to combinatorial explosion of the search space for finding complex patterns; a viable alternative is network modeling for abstraction and quantifying intrinsic data associations. Prominent network analysis methods furnish frameworks for model synthesis and validation but rely on standard correlation measures impaired by semi-supervised biases, latent heterogeneity, and uneven discretization techniques. Here we investigate a holistic measure for encapsulating data heterogeneity for enhanced efficacy of revealing complex patterns through network analysis. Our unique correlation metric, K-medoids Utility for Duo Original Similarities (Kudos), exhaustively factors real-valued analyte data to compute …

A Survey On Online Matching And Ad Allocation, Ryan Lee

Theses

One of the classical problems in graph theory is matching. Given an undirected graph, find a matching which is a set of edges without common vertices. In 1990s, Richard Karp, Umesh Vazirani, and Vijay Vazirani would be the first computer scientists to use matchings for online algorithms [8]. In our domain, an online algorithm operates in the online setting where a bipartite graph is given. On one side of the graph there is a set of advertisers and on the other side we have a set of impressions. During the online phase, multiple impressions will arrive and the objective of …

Accessible And Functional Visualizations For Exploring And Analyzing The Writing And Programming Process, Shaney Flores

Theses

Background. Developing easily decodable and insightful visuals is a challenge in the field of data visualization. This challenge becomes more pronounced when the data is of a complex nature. Good visualizations clearly convey their data using designs with appropriate encoding, visual attributes (i.e., color, shape, size, etc.), and accessibility features (e.g., distinctive colors for color-blind individu- als). One area where well-designed visualizations can make a significant impact is elucidating the learning process. Users ranging from self-taught individuals to students enrolled in coursework could use such visuals to detect problematic areas in their process of learning skills such as writing …

Efficient And Scalable Triangle Centrality Algorithms In The Arkouda Framework, Joseph Thomas Patchett

Theses

Graph data structures provide a unique challenge for both analysis and algorithm development. These data structures are irregular in that memory accesses are not known a priori and accesses to these structures tend to lack locality.

Despite these challenges, graph data structures are a natural way to represent relationships between entities and to exhibit unique features about these relationships. The network created from these relationships can create unique local structures that can describe the behavior between members of these structures. Graphs can be analyzed in a number of different ways including at a high level in community detection and at …

Un-Fair Trojan: Targeted Backdoor Attacks Against Model Fairness, Nicholas Furth

Theses

Machine learning models have been shown to be vulnerable against various backdoor and data poisoning attacks that adversely affect model behavior. Additionally, these attacks have been shown to make unfair predictions with respect to certain protected features. In federated learning, multiple local models contribute to a single global model communicating only using local gradients, the issue of attacks become more prevalent and complex. Previously published works revolve around solving these issues both individually and jointly. However, there has been little study on the effects of attacks against model fairness. Demonstrated in this work, a flexible attack, which we call Un-Fair …

Disaster Site Structure Analysis: Examining Effective Remote Sensing Techniques In Blue Tarpaulin Inspection, Madeline G. Miles

Theses

This thesis aimed to evaluate three methods of analyzing blue roofing tarpaulin (tarp) placed on homes in post natural disaster zones with remote sensing techniques by assessing the different methods- image segmentation, machine learning (ML), and supervised classification. One can determine which is the most efficient and accurate way of detecting blue tarps. The concept here was that using the most efficient and accurate way to locate blue tarps can aid federal, state, and local emergency management (EM) operations and homeowners. In the wake of a natural disaster such as a tornado, hurricane, thunderstorm, or similar weather events, roofs are …

The Role Of Physical Geographic Features In Video Games, Dexter Ferguson

Theses

This thesis examines how physical geographic features affect individuals when interacting with video game environments. There are very few studies on the geography of video games. I aim to bridge the gap between game studies and geography, a naturally occurring link. The observation of elements such as the geography of an area is critical when exploring an individual’s perception of a video game environment. Individuals interact with the geography of their location on a day-to-day basis, whether it be physical or human geography; how they can relate a video game’s geography to the real world dramatically affects their immersion. An …

Pranayama Breathing Detection With Deep Learning, Bikash Shrestha

Theses

Yoga, a complementary health approach, according to a 2017 National Health Interview Survey by the Center for Disease Control and Prevention (CDC), is a choice of around 14.3% adults in the US. Kapalbhati pranayama, a yoga practice of alternating fast exhales and longer passive inhales, is understood to improve our health. Incorrect and irregular practices, however, can cause injuries and adverse effects. To avoid these undesired effects, it is essential to maintain a pace fit for the practitioner. In the absence of any tools to observe a pace of practice, this work develops a deep learning method that listens to …

Stationary Probability Distributions Of Stochastic Gradient Descent And The Success And Failure Of The Diffusion Approximation, William Joseph Mccann

Theses

In this thesis, Stochastic Gradient Descent (SGD), an optimization method originally popular due to its computational efficiency, is analyzed using Markov chain methods. We compute both numerically, and in some cases analytically, the stationary probability distributions (invariant measures) for the SGD Markov operator over all step sizes or learning rates. The stationary probability distributions provide insight into how the long-time behavior of SGD samples the objective function minimum.

A key focus of this thesis is to provide a systematic study in one dimension comparing the exact SGD stationary distributions to the Fokker-Planck diffusion approximation equations —which are commonly used in …

Time Series Forecasting With Applications To Finance, Viswapriya Misra

Theses

In finance, many phenomena are modeled as time series. This thesis investigates time series forecasting problems in finance, precisely the stock price prediction problem. We employ and compare traditional statistical algorithms like MA, ARIMA, and ARMA-GARCH with newly developed deep learning-based algorithms such RNNs, LSTMs, GRUs, TCNs, and bidirectional LSTMs and GRUs for predicting stock prices. We perform a comprehensive study and present all the experimental results on different datasets. We find that ARIMA and GRU perform better for single-step stock price prediction than other deep learning architectures. Adding market and economic indicators do not improve the performance of the …

Short Term Temperature Forecasting Using Lstms, And Cnn, Darshan Shah

Theses

Weather forecasting is a vital application in present times. We can use the predictions to minimize the weather related loss. Use of machine learning and deep learning algorithms for forecasting, can eliminate or reduce the necessity of big data and high computation dependent process of parameterization. Long Short-Term Memory (LSTM) is a widely used deep learning architecture for time series forecasting. In this paper, we aim to predict one day ahead average temperature using a 2-layer neural network consisting of one layer of LSTM and one layer of 1D convolution. The input is pre-processed using a smoothing technique and output …

Team Formation Using Recommendation Systems, Shreyas Patil

Theses

The importance of team formation has been realized since ages, but finding the most effective team out of the available human resources is a problem that persists to the date. Having members with complementary skills, along with a few must-have behavioral traits, such as trust and collaborativeness among the team members are the key ingredients behind team synergy and performance. This thesis designs and implements two different algorithms for the team formation problem using ideas adapted from the recommender systems literature. One of the proposed solutions uses the Glicko-2 rating system to rate the employees’ skills which can easily separate …

Privacy-Preserving Recommendation System Using Federated Learning, Rahul Basu

Theses

Federated Learning is a form of distributed learning which leverages edge devices for training. It aims to preserve privacy by communicating users’ learning parameters and gradient updates to the global server during the training while keeping the actual data on the users’ devices. The training on global server is performed on these parameters instead of user data directly while fine tuning of the model can be done on client’s devices locally. However, federated learning is not without its shortcomings and in this thesis, we present an overview of the learning paradigm and propose a new federated recommender system framework that …

Semi-Automatic Management Of Knowledge Bases Using Formal Ontologies, Andreas Textor

Theses

This thesis presents an approach that deals with the ever-growing amount of data in knowledge bases, especially concerning knowledge interoperability and formal representation of domain knowledge. There arc multiple issues that must be addressed with current systems. A multitude of different formats, sources and tools exist in a domain, and it is desirable to develop their use further towards a standardised environment. Such an environment should support both the representation and processing of data from this domain, and the connection to other domains, where necessary. In order to manage large amounts of data, it should be possible to perform whatever …

Information Brokerage - A New Approach Using Knowledge Management, Robert Loew

Theses

Traditional knowledge management attempts to store the total enterprise knowledge into IT related structures, but knowledge management strategies should focus also on people and organisations and not just on technology.

This can be addressed in part by communicating knowledge between people, in and across organisations. Such communication focuses on the development and exchange of people’s experience.

This thesis introduces a hybrid knowledge management solution consisting of an automated system that includes people: employees, experts and Knowledge Brokers (KB). The hybrid concept supports the identification of suitable people for certain topics of discussion. Further, it supports the knowledge communication process by …