Physical Sciences and Mathematics Commons^™

Hierarchical Federated Learning On Healthcare Data: An Application To Parkinson's Disease, Brandon J. Harvill

Theses and Dissertations

Federated learning (FL) is a budding machine learning (ML) technique that seeks to keep sensitive data private, while overcoming the difficulties of Big Data. Specifically, FL trains machine learning models over a distributed network of devices, while keeping the data local to each device. We apply FL to a Parkinson’s Disease (PD) telemonitoring dataset where physiological data is gathered from various modalities to determine the PD severity level in patients. We seek to optimally combine the information across multiple modalities to assess the accuracy of our FL approach, and compare to traditional ”centralized” statistical and deep learning models.

Automated Registration Of Titanium Metal Imaging Of Aircraft Components Using Deep Learning Techniques, Nathan A. Johnston

Theses and Dissertations

Studies have shown a connection between early catastrophic engine failures with microtexture regions (MTRs) of a specific size and orientation on the titanium metal engine components. The MTRs can be identified through the use of Electron Backscatter Diffraction (EBSD) however doing so is costly and requires destruction of the metal component being tested. A new methodology of characterizing MTRs is needed to properly evaluate the reliability of engine components on live aircraft. The Air Force Research Lab Materials Directorate (AFRL/RX) proposed a solution of supplementing EBSD with two non-destructive modalities, Eddy Current Testing (ECT) and Scanning Acoustic Microscopy (SAM). Doing …

Drone Detection Using Yolov5, Burchan Aydin, Subroto Singha

Faculty Publications

The rapidly increasing number of drones in the national airspace, including those for recreational and commercial applications, has raised concerns regarding misuse. Autonomous drone detection systems offer a probable solution to overcoming the issue of potential drone misuse, such as drug smuggling, violating people’s privacy, etc. Detecting drones can be difficult, due to similar objects in the sky, such as airplanes and birds. In addition, automated drone detection systems need to be trained with ample amounts of data to provide high accuracy. Real-time detection is also necessary, but this requires highly configured devices such as a graphical processing unit (GPU). …

Improving Automatic Melanoma Diagnosis Using Deep Learning-Based Segmentation Of Irregular Networks, Anand K. Nambisan, Akanksha Maurya, Norsang Lama, Thanh Phan, Gehana Patel, Keith Miller, Binita Lama, Jason Hagerty, Ronald Stanley, William V. Stoecker

Chemistry Faculty Research & Creative Works

Deep Learning Has Achieved Significant Success in Malignant Melanoma Diagnosis. These Diagnostic Models Are Undergoing a Transition into Clinical Use. However, with Melanoma Diagnostic Accuracy in the Range of Ninety Percent, a Significant Minority of Melanomas Are Missed by Deep Learning. Many of the Melanomas Missed Have Irregular Pigment Networks Visible using Dermoscopy. This Research Presents an Annotated Irregular Network Database and Develops a Classification Pipeline that Fuses Deep Learning Image-Level Results with Conventional Hand-Crafted Features from Irregular Pigment Networks. We Identified and Annotated 487 Unique Dermoscopic Melanoma Lesions from Images in the ISIC 2019 Dermoscopic Dataset to Create a …

Revealing The Three-Dimensional Magnetic Texture With Machine Learning Models, Shihua Zhao

Dissertations, Theses, and Capstone Projects

Revealing three-dimensional (3D) magnetic textures with vector field electron tomography (VFET) is essential in studying novel magnetic materials with topologically protected spin textures potentially being used in the next-generation semiconductor industry. In this dissertation, we use machine learning (ML) models to reconstruct 3D magnetic textures from electron holography (EH) data.

We can feed the EH data, a series of two-dimensional (2D) phasemaps, into a neural network (NN) architecture directly or feed the EH data into a conventional VFET and then feed the reconstructed results into a NN. Thus, perceptive NN, either a simple convolutional neural network (CNN) or Unet architecture, …

Data Poisoning: A New Threat To Artificial Intelligence, Nary Simms

Mathematics and Computer Science Capstones

Artificial Intelligence (AI) adoption is rapidly being deployed in a number of fields, from banking and finance to healthcare, robotics, transportation, military, e-commerce and social networks. Grand View Research estimates that the global AI market was worth 93.5 billion in 2021 and that it will increase at a compound annual growth rate (CAGR) of 38.1% from 2022 to 2030. According to a 2020 MIT Sloan Management survey, 87% of multinational corporations believe that AI technology will provide a competitive edge. Artificial Intelligence relies heavily on datasets to train its models. The more data, the better it learns and predicts. However, …

Digital Archaeology: Detection Of Archaeological Structures Using Convolutional Neural Networks On Aerial Lidar Data, Katie Larue

WWU Honors College Senior Projects

Archaeology is a field that is mostly done by hand. Archaeologists explore remote and unknown areas of the world to find undiscovered civilizations that will give us any idea about how people lived in the past. To speed up this process, Airborne light detection and ranging or LiDAR systems have been used to great effect to speed up this processing. However, we still require domain experts to annotate this information to confirm structures. Deep learning has the potential to speed up this process and the following presentation is a basic overview of machine learning, popular types of deep learning models, …

Using Machine Learning For Web Accessibility, Tlamelo Makati

Academic Posters Collection

This research will explore the potential of machine learning to enhance web accessibility. Web accessibility is typically defined in terms of Web Accessibility Guidelines (WCAG), which states that everyone should be able to perceive, operate, understand and interpret the web regardless of disability or use of assistive technology. We would like to consult digital accessibility experts through interviews and focus groups to understand the web accessibility auditing and remediation processes in detail, with a focus on web navigation. An important goal of this work is to establish development processes where all stakeholders can leverage machine-learning tools to produce more accessible …

Adaptive Resolution Loss: An Efficient And Effective Loss For Time Series Self-Supervised Learning Framework, Kevin Garcia, Juan Manuel Perez, Yifeng Gao

Computer Science Faculty Publications and Presentations

Time series data is a crucial form of information that has vast opportunities. With the widespread use of sensor networks, largescale time series data has become ubiquitous. One of the most prominent problems in time series data mining is representation learning. Recently, with the introduction of self-supervised learning frameworks (SSL), numerous amounts of research have focused on designing an effective SSL for time series data. One of the current state-of-the-art SSL frameworks in time series is called TS2Vec. TS2Vec specially designs a hierarchical contrastive learning framework that uses loss-based training, which performs outstandingly against benchmark testing. However, the computational cost …

Background Discrimination Of A Neutrino Detector With Dense Neural Networks, Perry Siehien

Dissertations and Theses

Neutrinos are subatomic particles that weakly interact with matter due to their neutral charge and small cross section. Detectors that search for neutrinos require sensitive instrumentation, which makes them susceptible to various background sources such as gamma rays. Additionally, coherent elastic neutrino-nucleus scattering events, or CEvNS, are the weakest neutrino interactions at 1-25 keV, making them exceptionally difficult to observe. To understand the physics of CEvNS events within the detector material, the recoil signatures of relevant interactions must be determined. Traditional analysis methods are effective, but cannot be applied to energies below 50 keV, due to the overlap of discrimination …

Liquid Tab, Nathan Hulet

Williams Honors College, Honors Research Projects

Guitar transcription is a complex task requiring significant time, skill, and musical knowledge to achieve accurate results. Since most music is recorded and processed digitally, it would seem like many tools to digitally analyze and transcribe the audio would be available. However, the problem of automatic transcription presents many more difficulties than are initially evident. There are multiple ways to play a guitar, many diverse styles of playing, and every guitar sounds different. These problems become even more difficult considering the varying qualities of recordings and levels of background noise.

Machine learning has proven itself to be a flexible tool …

Analyzing Ground Motion Records With Cvi Fuzzy Art, Dustin Tanksley, Xinzhe Yuan, Genda Chen, Donald C. Wunsch

Civil, Architectural and Environmental Engineering Faculty Research & Creative Works

This paper explores using Cluster Validity Indices Fuzzy Adaptative Resonance Theory (CVI Fuzzy ART) to cluster ground motion records (GMRs). Clustering the features extracted from a supervised network trained for predicting the structure damage results in less overfitting from the trained network. Using Cluster Validity Indices (CVIs) to evaluate the clustering gives feedback to how well the data is being classified, allowing further separation of the data. By using CVI Fuzzy ART in combination with features extracted from a trained Convolutional Neural Network (CNN), we were able to form additional clusters in the data. Within the primary clusters, accuracy was …

Application Of Sentiment Analysis And Machine Learning Techniques To Predict Daily Cryptocurrency Price Returns, Edward Wu

CMC Senior Theses

This paper examines the effects of social media sentiment relating to Bitcoin on the daily price returns of Bitcoin and other popular cryptocurrencies by utilizing sentiment analysis and machine learning techniques to predict daily price returns. Many investors think that social media sentiment affects cryptocurrency prices. However, the results of this paper find that social media sentiment relating to Bitcoin does not add significant predictive value to forecasting daily price returns for each of the six cryptocurrencies used for analysis and that machine learning models that do not assume linearity between the current day price return and previous daily price …

On The Pursuit Of Developer Happiness: Webcam-Based Eye Tracking And Affect Recognition In The Ide, Tamsin Rogers

Honors Theses

Recent research highlights the viability of webcam-based eye tracking as a low-cost alternative to dedicated remote eye trackers. Simultaneously, research shows the importance of understanding emotions of software developers, where it was found that emotions have significant effects on productivity, code quality, and team dynamics. In this paper, we present our work towards an integrated eye-tracking and affect recognition tool for use during software development. This combined approach could enhance our understanding of software development by combining information about the code developers are looking at, along with the emotions they experience. The presented tool utilizes an unmodified webcam to capture …

Machine Learning Framework For Real-World Electronic Health Records Regarding Missingness, Interpretability, And Fairness, Jing Lucas Liu

Theses and Dissertations--Computer Science

Machine learning (ML) and deep learning (DL) techniques have shown promising results in healthcare applications using Electronic Health Records (EHRs) data. However, their adoption in real-world healthcare settings is hindered by three major challenges. Firstly, real-world EHR data typically contains numerous missing values. Secondly, traditional ML/DL models are typically considered black-boxes, whereas interpretability is required for real-world healthcare applications. Finally, differences in data distributions may lead to unfairness and performance disparities, particularly in subpopulations.

This dissertation proposes methods to address missing data, interpretability, and fairness issues. The first work proposes an ensemble prediction framework for EHR data with large missing …

Code Execution Capability As A Metric For Machine Learning–Assisted Software Vulnerability Detection Models, Daniel Grahn, Lingwei Chen, Junjie Zhang

Computer Science and Engineering Faculty Publications

In this paper, we consider how the ability to learn Code Execution Tasks affects a model’s accuracy on software vulnerability detection (SVD) benchmark datasets. We initially find that models can achieve near state-of-the-art accuracy on SVD benchmarks regardless of their ability to learn Code Execution Tasks. However, these models fail to generalize well across SVD benchmarks. The results indicate a bias in the datasets that allows models to predict non- SVD signals. Under the theory that different collection methods will reduce biases, we investigate combining the SVD datasets. When trained on combined datasets, SVD accuracy is reduced but correlation with …

Methods For Improving Potassium Fertilizer Recommendations For Corn In South Dakota, Andrew J. Ahlersmeyer

Electronic Theses and Dissertations

Corn (Zea mays L.) is a vital commodity in South Dakota’s agricultural sector. Optimal corn production occurs when there are sufficient mineral nutrients in the soil, especially potassium (K). Applications of K fertilizer are used when soil test K (STK) levels are deficient. Therefore, producers need reliable, thoroughly tested fertilizer recommendations to make profitable decisions and maintain environmental stewardship. South Dakota K fertilizer recommendations have not been updated in nearly 20 years. Simultaneously, changes in corn genetics, management practices, and climate patterns suggest that the critical soil test value (CSTV) for STK may have shifted in that same time frame. …

Breast Density Classification Using Deep Learning, Conrad Thomas Testagrose

UNF Graduate Theses and Dissertations

Breast density screenings are an accepted means to determine a patient's predisposed risk of breast cancer development. Although the direct correlation is not fully understood, breast cancer risk increases with higher levels of mammographic breast density. Radiologists visually assess a patient's breast density using mammogram images and assign a density score based on four breast density categories outlined by the Breast Imaging and Reporting Data Systems (BI-RADS). There have been efforts to develop automated tools that assist radiologists with increasing workloads and to help reduce the intra- and inter-rater variability between radiologists. In this thesis, I explored two deep-learning-based approaches …

A Symbolic Music Transformer For Real-Time Expressive Performance And Improvisation, Arnav Shirodkar

Senior Projects Fall 2023

With the widespread proliferation of AI technology, deep architectures — many of which are based on neural networks — have been incredibly successful in a variety of different research areas and applications. Within the relatively new domain of Music Information Retrieval (MIR), deep neural networks have also been successful for a variety of tasks, including tempo estimation, beat detection, genre classification, and more. Drawing inspiration from projects like George E. Lewis's Voyager and Al Biles's GenJam, two pioneering endeavors in human-computer interaction, this project attempts to tackle the problem of expressive music generation and seeks to create a Symbolic Music …

Leveraging A Machine Learning Based Predictive Framework To Study Brain-Phenotype Relationships, Sage Hahn

Graduate College Dissertations and Theses

An immense collective effort has been put towards the development of methods forquantifying brain activity and structure. In parallel, a similar effort has focused on collecting experimental data, resulting in ever-growing data banks of complex human in vivo neuroimaging data. Machine learning, a broad set of powerful and effective tools for identifying multivariate relationships in high-dimensional problem spaces, has proven to be a promising approach toward better understanding the relationships between the brain and different phenotypes of interest. However, applied machine learning within a predictive framework for the study of neuroimaging data introduces several domain-specific problems and considerations, leaving the …

Developing Muscle Synergy Functions For Remote Gait Analysis, Nicole Marie Donahue

Graduate College Dissertations and Theses

Digital medicine promises to improve healthcare and enable its delivery to rural and underserved communities. A key component of digital medicine is accurate and robust remote patient monitoring. For example, remote monitoring of biomechanical measures of limb impairment during daily life could allow near real-time tracking of rehabilitation progress and personalization of rehabilitation paradigms in those recovering from orthopedic surgery. Wearable sensors have long been suggested as a means for quantifying muscle and joint loading, which can provide a direct measure of limb impairment. However, current approaches either do not provide these measures or require unwieldy wearable sensor arrays and/or …

Biomarker Identification For Breast Cancer Types Using Feature Selection And Explainable Ai Methods, David E. La Rosa Giraud

Honors Undergraduate Theses

This paper investigates the impact the LASSO, mRMR, SHAP, and Reinforcement Feature Selection techniques on random forest models for the breast cancer subtypes markers ER, HER2, PR, and TN as well as identifying a small subset of biomarkers that could potentially cause the disease and explain them using explainable AI techniques. This is important because in areas such as healthcare understanding why the model makes a specific decision is important it is a diagnostic of an individual which requires reliable AI. Another contribution is using feature selection methods to identify a small subset of biomarkers capable of predicting if a …

Utilizing Machine Learning In Healthcare In An Ethical Fashion, Nishka Ayyar

CMC Senior Theses

This thesis paper explores the ethical considerations surrounding the use of machine learning (ML) solutions in healthcare. The background section discusses the basics of machine learning techniques and algorithms, and the increasing interest in their utilization in the healthcare sector. The paper then reviews and critically analyzes four studies that highlight concerns related to using ML in healthcare, including issues of bias, privacy, accountability, and transparency. Based on the analysis of these studies, the paper presents several recommendations for addressing these concerns. The paper concludes with a discussion on the potential benefits of using machine learning technology in healthcare. Ultimately, …

Unsupervised-Based Distributed Machine Learning For Efficient Data Clustering And Prediction, Vishnu Vardhan Baligodugula

Browse all Theses and Dissertations

Machine learning techniques utilize training data samples to help understand, predict, classify, and make valuable decisions for different applications such as medicine, email filtering, speech recognition, agriculture, and computer vision, where it is challenging or unfeasible to produce traditional algorithms to accomplish the needed tasks. Unsupervised ML-based approaches have emerged for building groups of data samples known as data clusters for driving necessary decisions about these data samples and helping solve challenges in critical applications. Data clustering is used in multiple fields, including health, finance, social networks, education, and science. Sequential processing of clustering algorithms, like the K-Means, Minibatch K-Means, …

Towards A Novel Approach For Smart Agriculture Predictability, Rima Grati, Myriam Aloulou, Khouloud Boukadi

All Works

No abstract provided.

Explainable Machine Learning For Evapotranspiration Prediction, Bamory Koné, Rima Grati, Bassem Bouaziz, Khouloud Boukadi

All Works

No abstract provided.

Analysis Of Chemical Elements In Basalts Using Mislabeled Data, A Machine Learning Approach, Jenifer Vivar

Dissertations and Theses

Scientists use basalt chemistry to discriminate among different tectonic settings. There are well-known chemical elements used to classify tectonic settings. An exploration of new features is done using Logistic Regression and Random Forest to discover any new elements of interest. The models were used with other tools, such as recursive feature elimination and permutations, to increase reliability. Among the scarcely explored chemical elements are Terbium (Tb), Holmium (Ho), Samarium (Sm), and Erbium (Er). The data used for the exploration contained many outliers. Therefore, an ensemble model was created to explore the location and composition of such outliers. The ensemble was …

Optimization Of Optical Nanosensor Response For The Detection Of Anthracyclines Using A Binary Machine Learning Classifier, Myesha Thahsin

Dissertations and Theses

Pharmacokinetic variables such as interindividual variation in metabolizing and eliminating drugs makes dose selection of chemotherapeutic anthracyclines difficult. One potential solution to determining dosing levels of an anthracycline is the development of non-invasive sensors to monitor their pharmacology in vivo. Single-walled carbon nanotubes (SWCNT) have substantial potential for in vivo sensor development, as they exhibit near-infrared fluorescence in the tissue-transparent window and a robust response to their local environment. An emerging method for evaluating and optimizing SWCNT sensor response is through machine learning. In this study, anthracyclines Daunorubicin, Doxorubicin, Epirubicin, Mitoxantrone and Idarubicin, were used to interrogate 12 SWCNT preparations …

Predicting Housing Prices Using Ai, Eric Sconyers

Williams Honors College, Honors Research Projects

I have created an AI model that can predict housing prices with 70 percent accuracy in Ames Iowa. I was able to use data from a website called Kaggle.com which is a website that provides datasets to the public so they can create AI models with the data. I found the dataset pertaining to housing prices in Ames Iowa. With this data, I was able to create an AI model that can predict the housing price of these homes. The technology I used in this project was Python as the programming language, and I used the scikit-learn library which has …

Human Tracking Function For Robotic Dog, Andrew Sharkey

Williams Honors College, Honors Research Projects

With the increase the increase in automation and humans and robots working side by side, there is a need for a more organic way of controlling robots. The goal of this project is to create a control system for Boston dynamics robotic dog Spot that implements human tracking image software to follow humans using computer vision as well as using hand tracking image software to allow for control input through hand gestures.