Physical Sciences and Mathematics Commons^™

Gt-Ches And Dycon: Improved Classification For Human Evolutionary Systems, Joseph S. Johnson

Theses and Dissertations

The purpose of this work is to rethink the process of learning in human evolutionary systems. We take a sober look at how game theory, network theory, and chaos theory pertain specifically to the modeling, data, and training components of generalization in human systems. The value of our research is three-fold. First, our work is a direct approach to align machine learning generalization with core behavioral theories. We made our best effort to directly reconcile the axioms of these heretofore incompatible disciplines -- rather than moving from AI/ML towards the behavioral theories while building exclusively on AI/ML intuition. Second, this …

Adaptive Multi-Label Classification On Drifting Data Streams, Martha Roseberry

Theses and Dissertations

Drifting data streams and multi-label data are both challenging problems. When multi-label data arrives as a stream, the challenges of both problems must be addressed along with additional challenges unique to the combined problem. Algorithms must be fast and flexible, able to match both the speed and evolving nature of the stream. We propose four methods for learning from multi-label drifting data streams. First, a multi-label k Nearest Neighbors with Self Adjusting Memory (ML-SAM-kNN) exploits short- and long-term memories to predict the current and evolving states of the data stream. Second, a punitive k nearest neighbors algorithm with a self-adjusting …

Towards Machine Learning-Based Control Of Autonomous Vehicles In Solar Panel Cleaning Systems, Farima Hajiahmadi

Theses and Dissertations

This thesis presents a machine learning (ML)-based approach for the intelligent control of Autonomous Vehicles (AVs) utilized in solar panel cleaning systems, aiming to mitigate challenges arising from uncertainties, disturbances, and dynamic environments. Solar panels, predominantly situated in dedicated lands for solar energy production (e.g., agricultural solar farms), are susceptible to dust and debris accumulation, leading to diminished energy absorption. Instead of labor-intensive manual cleaning, robotic cleaners offer a viable solution. AVs equipped to transport and precisely position these cleaning robots are indispensable for efficient navigation among solar panel arrays. However, environmental obstacles (e.g., rough terrain), variations in solar panel …

Phenotyping Cotton Compactness Using Machine Learning And Uas Multispectral Imagery, Joshua Carl Waldbieser

Theses and Dissertations

Breeding compact cotton plants is desirable for many reasons, but current research for this is restricted by manual data collection. Using unmanned aircraft system imagery shows potential for high-throughput automation of this process. Using multispectral orthomosaics and ground truth measurements, I developed supervised models with a wide range of hyperparameters to predict three compactness traits. Extreme gradient boosting using a feature matrix as input was able to predict the height-related metric with R2=0.829 and RMSE=0.331. The breadth metrics require higher-detailed data and more complex models to predict accurately.

Accelerating Machine Learning Inference For Satellite Component Feature Extraction Using Fpgas., Andrew Ekblad

Theses and Dissertations

Running computer vision algorithms requires complex devices with lots of computing power, these types of devices are not well suited for space deployment. The harsh radiation environment and limited power budgets have hindered the ability of running advanced computer vision algorithms in space. This problem makes running an on-orbit servicing detection algorithm very difficult. This work proposes using a low powered FPGA to accelerate the computer vision algorithms that enable satellite component feature extraction. This work uses AMD/Xilinx’s Zynq SoC and DPU IP to run model inference. Experiments in this work centered around improving model post processing by creating implementations …

Improving Xrd Analysis With Machine Learning, Rachel E. Drapeau

Theses and Dissertations

X-ray diffraction analysis (XRD) is an inexpensive method to quantify the relative proportions of mineral phases in a rock or soil sample. However, the analytical software available for XRD requires extensive user input to choose phases to include in the analysis. Consequently, analysis accuracy depends greatly on the experience of the analyst, especially as the number of phases in a sample increases (Raven & Self, 2017; Omotoso, 2006). The purpose of this project is to test whether incorporating machine learning methods into XRD software can improve the accuracy of analyses by assisting in the phase-picking process. In order to provide …

Tornado Outbreak False Alarm Probabilistic Forecasts With Machine Learning, Kirsten Reed Snodgrass

Theses and Dissertations

Tornadic outbreaks occur annually, causing fatalities and millions of dollars in damage. By improving forecasts, the public can be better equipped to act prior to an event. False alarms (FAs) can hinder the public’s ability (or willingness) to act. As such, a probabilistic FA forecasting scheme would be beneficial to improving public response to outbreaks.

Here, a machine learning approach is employed to predict FA likelihood from Storm Prediction Center (SPC) tornado outbreak forecasts. A database of hit and FA outbreak forecasts spanning 2010 – 2020 was developed using historical SPC convective outlooks and the SPC Storm Reports database. Weather …

Self-Supervised Representation Learning For Motion Time Series: A Case Study In Activity Recognition, Luis Carlos Garza Perez

Theses and Dissertations

In this thesis we will learn about what contrastive learning and time series are and understand the differences between supervised and self-supervised frameworks in machine learning. In addition, we will describe how the newest and most efficient self-supervised learning framework for visual representations to this date works, called SimCLR, which was originally developed to obtain useful vector representations from static images. We will also explain what TS2Vec is, and how a combination of both approaches can be applied to the concept of a time series, and still be able to extract a vector representation of the subject described by the …

A Survey Of Graph Neural Networks On Synthetic Data, Brigham Stone Carson

Theses and Dissertations

We relate properties of attributed random graph models to the performance of GNN architectures. We identify regimes where GNNs outperform feedforward neural networks and non-attributed graph clustering methods. We compare GNN performance on our synthetic benchmark to performance on popular real-world datasets. We analyze the theoretical foundations for weak recovery in GNNs for popular one- and two-layer architectures. We obtain an explicit formula for the performance of a 1-layer GNN, and we obtain useful insights on how to proceed in the 2-layer case. Finally, we improve the bound for a notable result on the GNN size generalization problem by 1.

Language Modeling Using Image Representations Of Natural Language, Seong Eun Cho

Theses and Dissertations

This thesis presents training of an end-to-end autoencoder model using the transformer, with an encoder that can encode sentences into fixed-length latent vectors and a decoder that can reconstruct the sentences using image representations. Encoding and decoding sentences to and from these image representations are central to the model design. This method allows new sentences to be generated by traversing the Euclidean space, which makes vector arithmetic possible using sentences. Machines excel in dealing with concrete numbers and calculations, but do not possess an innate infrastructure designed to help them understand abstract concepts like natural language. In order for a …

Machine Learning Models Interpretability For Malware Detection Using Model Agnostic Language For Exploration And Explanation, Ikuromor Mabel Ogiriki

Theses and Dissertations

The adoption of the internet as a global platform has birthed a significant rise in cyber-attacks of various forms ranging from Trojans, worms, spyware, ransomware, botnet malware, rootkit, etc. In order to tackle the issue of all these forms of malware, there is a need to understand and detect them. There are various methods of detecting malware which include signature, behavioral, and machine learning. Machine learning methods have proven to be the most efficient of all for malware detection. In this thesis, a system that utilizes both the signature and dynamic behavior-based detection techniques, with the added layer of the …

Atomlbs: An Atom Based Convolutional Neural Network For Druggable Ligand Binding Site Prediction, Md Ashraful Islam

Theses and Dissertations

Despite advances in drug research and development, there are few and ineffective treatments for a variety of diseases. Virtual screening can drastically reduce costs and accelerate the drug discovery process. Binding site identification is one of the initial and most important steps in structure-based virtual screening. Identifying and defining protein cavities that are likely to bind to a small compound is the objective of this task. In this research, we propose four different convolutional neural networks for predicting ligand-binding sites in proteins. A parallel optimized data pipeline is created to enable faster training of these neural network models on minimal …

Cnn-Based Dendrite Core Detection From Microscopic Images Of Directionally Solidified Ni-Base Alloys, Xiaoguang Li

Theses and Dissertations

Dendrite core is the center point of the dendrite. The information of dendrite core is very helpful for material scientists to analyze the properties of materials. Therefore, detecting the dendrite core is a very important task in the material science field. Meanwhile, because of some special properties of the dendrites, this task is also very challenging. Different from the typical detection problems in the computer vision field, detecting the dendrite core aims to detect a single point location instead of the bounding-box. As a result, the existing regressing bounding-box based detection methods can not work well on this task because …

Search For Triple-Proton Decay Using Machine Learning With Cuore, Douglas Adams

Theses and Dissertations

A framework to search for a triple-proton decay of ¹³⁰Te in the CUORE detector against a background of muons is presented. We use machine learning to classify different kinds of energy depositing events. We use the classification information to improve our detection or non-detection limits of a triple-proton decay process. We derive and use a methodology of combining Poisson counting statistics with supervised classification machine learning tools. Additionally, a sensitivity calculation is provided which uses the classification counting likelihood. Using our analysis technique, we achieve an lower 2σ half-life bound of 7.43×10²⁴yrs for triple-proton decay of …

Enabling Rapid Chemical Analysis Of Plutonium Alloys Via Machine Learning-Enhanced Atomic Spectroscopy Techniques, Ashwin P. Rao

Theses and Dissertations

Analytical atomic spectroscopy methods have the potential to provide solutions for rapid, high fidelity chemical analysis of plutonium alloys. Implementing these methods with advanced analytical techniques can help reduce the chemical analysis time needed for plutonium pit production, directly enabling the 80 pit-per-year by 2030 manufacturing goal outlined in the 2018 Nuclear Posture Review. Two commercial, handheld elemental analyzers were validated for potential in situ analysis of Pu. A handheld XRF device was able to detect gallium in a Pu surrogate matrix with a detection limit of 0.002 wt% and a mean error of 8%. A handheld LIBS device was …

Leveraging Subject Matter Expertise To Optimize Machine Learning Techniques For Air And Space Applications, Philip Y. Cho

Theses and Dissertations

We develop new machine learning and statistical methods that are tailored for Air and Space applications through the incorporation of subject matter expertise. In particular, we focus on three separate research thrusts that each represents a different type of subject matter knowledge, modeling approach, and application. In our first thrust, we incorporate knowledge of natural phenomena to design a neural network algorithm for localizing point defects in transmission electron microscopy (TEM) images of crystalline materials. In our second research thrust, we use Bayesian feature selection and regression to analyze the relationship between fighter pilot attributes and flight mishap rates. We …

Determining The Effects Of Elevated Carbon Dioxide On Soil Acidification, Cation Depletion, And Soil Inorganic Carbon And Mapping Soil Carbons Using Artificial Intelligence, Jannatul Ferdush

Theses and Dissertations

Soil carbon is the largest sink and source of the global carbon cycle and is disturbed by several natural, anthropogenic, and environmental factors. The global increase of atmospheric CO₂ affects soil carbon cycling through varied biogeochemical processes. The first chapter is a compilation of current information on potential factors triggering soil acidification and weathering mechanisms under elevated CO₂ and their consequences on soil inorganic carbon (SIC) pool and quality. Soil water content and precipitation were critical factors influencing elevated CO₂ effects on the SIC pool. The second chapter examines a detailed column experiment in which six soils …

Classification Models For 2,4-D Formulations In Damaged Enlist Crops Through The Application Of Ftir Spectroscopy And Machine Learning Algorithms, Benjamin Blackburn

Theses and Dissertations

With new 2,4-Dichlorophenoxyacetic acid (2,4-D) tolerant crops, increases in off-target movement events are expected. New formulations may mitigate these events, but standard lab techniques are ineffective in identifying these 2,4-D formulations. Using Fourier-transform infrared spectroscopy and machine learning algorithms, research was conducted to classify 2,4-D formulations in treated herbicide-tolerant soybeans and cotton and observe the influence of leaf treatment status and collection timing on classification accuracy. Pooled Classification models using k-nearest neighbor classified 2,4-D formulations with over 65% accuracy in cotton and soybean. Tissue collected 14 DAT and 21 DAT for cotton and soybean respectively produced higher accuracies than the …

Using Machine Learning To Classify Volleyball Jumps, Miki Jauhiainen

Theses and Dissertations

In this study, inertial measurement units (IMUs) were used to train a random forest classifier to correctly classify different jump types in volleyball. Athlete motion data were collected in a controlled setting using three IMUs, one on the waist and one on each ankle. There were 11 participants who at the time played volleyball at the collegiate level in the United States, seven male and four female. Each performed the same number of jumps across the eight jump types--five BASIC jumps and three each of the other seven--resulting in 26 jumps per subject for a total of 286. The data …

Reconstructing Historical Earthquake-Induced Tsunamis: Case Study Of 1820 Event Near South Sulawesi, Indonesia, Taylor Jole Paskett

Theses and Dissertations

We build on the method introduced by Ringer, et al., applying it to an 1820 event that happened near South Sulawesi, Indonesia. We utilize other statistical models to aid our Metropolis-Hastings sampler, including a Gaussian process which informs the prior. We apply the method to multiple possible fault zones to determine which fault is the most likely source of the earthquake and tsunami. After collecting nearly 80,000 samples, we find that between the two most likely fault zones, the Walanae fault zone matches the anecdotal accounts much better than Flores. However, to support the anecdotal data, both samplers tend toward …

Image-Based Crack Detection By Extracting Depth Of The Crack Using Machine Learning, Nishat Tabassum

Theses and Dissertations

Concrete structures have been a major aspect of social infrastructure since the ancient Roman times, so they have been used for many centuries. Concrete is used for the durability and support it provides to buildings and bridges. Assessing the state of these structures is important in preserving the longevity of structures and the safety of the public. Detecting cracks in their early stage allows repairs to be made without the need to replace the whole structure, so it reduces the cost. Traditional methods are slowly falling behind as technology advances and an increase in demand for a practical method of …

Language Learning Using Models Of Intentionality In Repeated Games With Cheap Talk, Jonathan Berry Skaggs

Theses and Dissertations

Language is critical to establishing long-term cooperative relationships among intelligent agents (including people), particularly when the agents' preferences are in conflict. In such scenarios, an agent uses speech to coordinate and negotiate behavior with its partner(s). While recent work has shown that neural language modeling can produce effective speech agents, such algorithms typically only accept previous text as input. However, in relationships among intelligent agents, not all relevant context is expressed in conversation. Thus, in this paper, we propose and analyze an algorithm, called Llumi, that incorporates other forms of context to learn to speak in long-term relationships modeled as …

Hardware Isolation Approach To Securely Use Untrusted Gpus In Cloud Environments For Machine Learning, Lucas D. Hall

Theses and Dissertations

Machine Learning (ML) is now a primary method for getting useful information out of the immense volumes of data being generated and stored in society today. Useful data is a commodity for training ML models and those that need data for training are often not the owners of the data leading to a desire to use cloud-based services. Deep learning algorithms are best suited to run on a graphical processing unit (GPU) which presents a specific problem since the GPU is not a secure or trusted piece of hardware in the cloud computing environment.

In this paper, we will analyze …

Physics-Guided Machine Learning In Ocean Acoustics Using Fisher Information, Michael Craig Mortenson

Theses and Dissertations

Waterborne acoustic signals carry information about the ocean environment. Ocean geoacoustic inversion is the task of estimating environmental parameters from received acoustic signals by matching the measured sound with the predictions of a physics-based model. A lower bound on the uncertainty associated with environmental parameter estimates, the Cramér-Rao bound, can be calculated from the Fisher information, which is dependent on derivatives of a physics-based model. Physics-based preconditioners circumvent the need for variable step sizes when computing numerical derivatives. This work explores the feasibility of using a neural network to perform geoacoustic inversion for environmental parameters and their associated uncertainties from …

Using Connections To Make Predictions On Dynamic Networks, Rebecca Dorff Jones

Theses and Dissertations

Networks are sets of objects that are connected in some way and appear abundantly in nature, sociology, and technology. For many centuries, network theory focused on static networks, which are networks that do not change. However, since all networks transform over time, static networks have limited applications. By comparison, dynamic networks model how connections between objects change over time. In this work, we will explore how connections in dynamic networks change and how we can leverage these changes to make predictions about future iterations of networks. We will do this by first considering the link prediction problem, using either Katz …

Telemetry Data Mining For Unmanned Aircraft Systems, Li Yu

Theses and Dissertations

With ever more data becoming available to the US Air Force, it is vital to develop effective methods to leverage this strategic asset. Machine learning (ML) techniques present a means of meeting this challenge, as these tools have demonstrated successful use in commercial applications. For this research, three ML methods were applied to a unmanned aircraft system (UAS) telemetry dataset with the aim of extracting useful insight related to phases of flight. It was shown that ML provides an advantage in exploratory data analysis and as well as classification of phases. Neural network models demonstrated the best performance with over …

Analysis Of Generalized Artificial Intelligence Potential Through Reinforcement And Deep Reinforcement Learning Approaches, Jonathan Turner

Theses and Dissertations

Artificial Intelligence is the next competitive domain; the first nation to develop human level artificial intelligence will have an impact similar to the development of the atomic bomb. To maintain the security of the United States and her people, the Department of Defense has funded research into the development of artificial intelligence and its applications. This research uses reinforcement learning and deep reinforcement learning methods as proxies for current and future artificial intelligence agents and to assess potential issues in development. Agent performance were compared across two games and one excursion: Cargo Loading, Tower of Hanoi, and Knapsack Problem, respectively. …

Improving Anonymized Search Relevance With Natural Language Processing And Machine Learning, Niko A. Petrocelli

Theses and Dissertations

Users often sacrifice personal data for more relevant search results, presenting a problem to communities that desire both search anonymity and relevant results. To balance these priorities, this research examines the impact of using Siamese networks to extend word embeddings into document embeddings and detect similarities between documents. The predicted similarity can locally re-rank search results provided from various sources. This technique is leveraged to limit the amount of information collected from a user by a search engine. A prototype is produced by applying the methodology in a real-world search environment. The prototype yielded an additional function of finding new …

Smoothing Of Convolutional Neural Network Classifications, Glen R. Drumm

Theses and Dissertations

Smoothing convolutional neural networks is investigated. When intermittent and random false predictions happen, a technique of average smoothing is applied to smooth out the incorrect predictions. While a simple problem environment shows proof of concept, obstacles remain for applying such a technique to a more operationally complex problem.

Incorporating Armed Escorts To The Military Medical Evacuation Dispatching Problem Via Stochastic Optimization And Reinforcement Learning, Andrew G. Gelbard

Theses and Dissertations

The military medical evacuation (MEDEVAC) dispatching problem seeks to determine high-quality dispatching policies to maximize the survivability of casualties within contingency operations. This research leverages applied operations research and machine learning techniques to solve the MEDEVAC dispatching problem and evaluate system performance. More specifically, we develop an infinite-horizon, continuous-time Markov decision process (MDP) model and approximate dynamic programming (ADP) solution approach to generate high-quality policies. The ADP solution approach utilizes an approximate value iteration algorithm strategy incorporating gradient descent Q-learning to approximate the value function. A notional, synthetically-generated scenario in Africa based around the capital city of Niger, Niamey is …