Digital Commons Network^™

The Generation Of A Physics Informed Machine Learning Model To Predict Defect Evolution In Materials & On The Thermally Activated Regime Of Dislocation Motion: A Simulation Driven Study On The Mechanical Behavior Of Crystals, Liam Myhill

All Theses

Line defects in crystals, known as dislocations, govern the mechanisms of plastic deformation at the micro-meso scale. The study of dislocations has proliferated the field of materials science and engineering for since the 1950’s, and modern studies show increasing utilization of computational methods to model the evolution of line defects in material systems. In keeping with modern research practice, the studies herewith demonstrate the use of advanced computing to generate models which can be used to better understand the behaviors of dislocations within crystal matrices. An advanced high-throughput model for a physically informed machine learning graph neural network (PIML-GNN) is …

Performance Based Design And Machine Learning In Structural Fire Engineering: A Case For Masonry, Deanna Craig

All Theses

The volatile and extreme nature of fire makes structural fire engineering unique in that the load actions dictating design are intense but not geographically or seasonally bound. Simply, fire can break out anywhere, at any time, and for any number of reasons. Despite the apparent need, fire design of structures still relies on expensive fire tests, complex finite element simulations, and outdated procedures with little room for innovation. This thesis will make a case for adopting the principles of performance-based design and machine learning in structural fire engineering to simplify the process and promote the consideration of fire in all …

State-Based Biological Communication, Nathan Clement

All Theses

Allostery (1) is the process through which proteins self-regulate in response to various stimuli. Allosteric interactions occur between nonadjacent spatially distant residues (1), and they are exhibited through the correlated motions (2) and momenta of participating residues. The location of allosteric sites in proteins can be determined experimentally but computational methods to predict the location of allosteric sites are being developed as well (2-4, 10). Experimental and computational methodologies for locating allosteric sites can be used to design specific targeted drug delivery (5-6, 19), but these methods have not yet …

A Quantitative Comparison Of Algorithmic And Machine Learning Network Flow Throughput Prediction, Cayden Wagner

All Theses

Applications ranging from video meetings, live streaming, video games, autonomous vehicle operations, and algorithmic trading heavily rely on low latency communication to operate optimally. A solution to fully support this growing demand for low latency is called dual-queue active queue management (AQM). Dual-queue AQM's functionality is reduced without network traffic throughput prediction.

Perhaps due to the current popularity of machine learning, there is a trend to adopt machine learning models over traditional algorithmic throughput prediction approaches without empirical support. This study tested the effectiveness of machine learning as compared to time series forecasting algorithms in predicting per-flow network traffic throughput …

Visualizing Features From Deep Neural Networks Trained On Alzheimer’S Disease And Few-Shot Learning Models For Alzheimer’S Disease, John Reeder

All Theses

Alzheimer’s disease is an incurable neural disease, usually affecting the elderly. The afflicted suffer from cognitive impairments that get dramatically worse at each stage. Previous research on Alzheimer’s disease analysis in terms of classification leveraged statistical models such as support vector machines. However, statistical models such as support vector machines train the from numerical data instead of medical images. Today, convolutional neural networks (CNN) are widely considered as the one which can achieve the state-of-the- art image classification performance. However, due to their black box nature, there can be reluctance amongst medical professionals for their use. On the other hand, …

Analysis Of Deep Learning Methods For Wired Ethernet Physical Layer Security Of Operational Technology, Lucas Torlay

All Theses

The cybersecurity of power systems is jeopardized by the threat of spoofing and man-in-the-middle style attacks due to a lack of physical layer device authentication techniques for operational technology (OT) communication networks. OT networks cannot support the active probing cybersecurity methods that are popular in information technology (IT) networks. Furthermore, both active and passive scanning techniques are susceptible to medium access control (MAC) address spoofing when operating at Layer 2 of the Open Systems Interconnection (OSI) model. This thesis aims to analyze the role of deep learning in passively authenticating Ethernet devices by their communication signals. This method operates at …

Diagnosis Of Myocardial Hypertrophic Disease States Through Machine Learning And Mechanistic Modeling, Michael William Ward

All Theses

Chronic pressure overload (PO) due to arterial hypertension can lead to structural changes within the heart including left ventricular hypertrophy (LVH) and eventually diastolic heart failure (DHF). The initial diagnosis of PO and LVH is typically challenging and costly, and thus, a new predictive diagnostic tool is desired. In a recent paper by Zile et al., it was found, through the use of a simple multivariate logistic regression, that there exists a multi-biomarker panel with predictive capabilities for the classification of patients with LVH and DHF. In our new work, we furthered the investigation into the plasma biomarker panel proposed …

Privacy-Preserving Image Classification Using Convolutional Neural Networks, David Karl Langbehn

All Theses

The process of image classification using convolutional neural networks (CNNs) often relies on access to large, annotated datasets and the use of cluster or cloud-based computing resources. However, many classification applications such as those in healthcare or defense introduce privacy concerns that prevent the collection of such data and the use of pre-existing large scale computing systems. Although many solutions to privacy preserving machine learning have previously been explored, the added computational complexity incurred with training on encrypted values inhibits these systems from executing in real-time. One of the most promising solutions that facilitates secure machine learning is secure multi-party …

Detection Of Delaminations In Carbon Fiber Reinforced Polymers Embedded With Terfenol-D Particles Using Machine Learning, Christopher Nelon

All Theses

The characterization of the damage state of a system provides insight into its performance and safety during operation. In composite materials, specifically, fiber-reinforced polymers, delaminations form from the evolution of cracks in a matrix that leads to adhesion failure between adjacent laminae. Nondestructive evaluation (NDE) seeks to characterize the state of a material or system during non-operational times. A previously proposed NDE method employs embedded magnetostrictive particles between laminae of carbon fiber reinforced polymer (CFRP) for damage sensing. The phenomenon of magnetostriction couples the mechanical state of a material with its magnetic state so that a change in the local …

A Comparison Of Machine Learning And Traditional Demand Forecasting Methods, Franz Stoll

All Theses

Obtaining accurate forecasts has been a challenging task to achieve for many organizations, both public and private. Today, many firms choose to share their internal information with supply chain partners to increase planning efficiency and accuracy in the hopes of making appropriate critical decisions. However, forecast errors can still increase costs and reduce profits. As company datasets likely contain both trend and seasonal behavior, this motivates the need for computational resources to find the best parameters to use when forecasting their data. In this thesis, two industrial datasets are examined using both traditional and machine learning (ML) forecasting methods. The …

A Machine Learning Framework For Energy Consumption Prediction, Chakara Rajan Madhusudanan

All Theses

Energy needs to be used very efficiently in today's world. With fast paced improvements in the industrial sector, demand is increasing, and energy efficiency programs become vital to reduce the energy wastage while also meeting the demand. The analysis of several scenarios used by policy makers suggest that for the global temperature to raise by less than 2° C by the end of this century, it is necessary to reduce industrial energy consumption increase by at least a half. To be on track with these scenarios and to achieve the desirable targets, it is important that we incorporate a dependable …

Designing Approximate Computing Circuits With Scalable And Systematic Data-Driven Techniques, Ling Qiu

All Theses

Semiconductor feature size has been shrinking significantly in the past decades. This decreasing trend of feature size leads to faster processing speed as well as lower area and power consumption. Among these attributes, power consumption has emerged as the primary concern in the design of integrated circuits in recent years due to the rapid increasing demand of energy efficient Internet of Things (IoT) devices. As a result, low power design approaches for digital circuits have become of great attractive in the past few years. To this end, approximate computing in hardware design has emerged as a promising design technique. It …

Compressing Deep Neural Networks Via Knowledge Distillation, Ankit Kulshrestha

All Theses

There has been a continuous evolution in deep neural network architectures since Alex Krizhevsky proposed AlexNet in 2012. Part of this has been due to increased complexity of the data and easier availability of datasets and part of it has been due to increased complexity of applications. These two factors form a self sustaining cycle and thereby have pushed the boundaries of deep learning to new domains in recent years.

Many datasets have been proposed for different tasks. In computer vision, notable datasets like ImageNet, CIFAR-10, 100, MS-COCO provide large training data, with different tasks like classification, segmentation and object …

Understanding 1d Convolutional Neural Networks Using Multiclass Time-Varying Signals, Ravisutha Sakrepatna Srinivasamurthy

All Theses

In recent times, we have seen a surge in usage of Convolutional Neural Networks to solve all kinds of problems - from handwriting recognition to object recognition and from natural language processing to detecting exoplanets. Though the technology has been around for quite some time, there is still a lot of scope to do research on what’s really happening ’under the hood’ in a CNN model.

CNNs are considered to be black boxes which learn something from complex data and provides desired results. In this thesis, an effort has been made to explain what exactly CNNs are learning by training …

Knowledge Extraction From Work Instructions Through Text Processing And Analysis, Abhiram Koneru

All Theses

The objective of this thesis is to design, develop and implement an automated approach to support processing of historical assembly data to extract useful knowledge about assembly instructions and time studies to facilitate the development of decision support systems, for a large automotive original equipment manufacturer (OEM). At a conceptual level, this research establishes a framework for sustainable and scalable approach to extract knowledge from big data using techniques from Natural Language Processing (NLP) and Machine Learning (ML). Process sheets are text documents that contain detailed instructions to assemble a portion of the vehicle, specification of parts and tools to …

Data-Intensive Computing For Bioinformatics Using Virtualization Technologies And Hpc Infrastructures, Pengfei Xuan

All Theses

The bioinformatics applications often involve many computational components and massive data sets, which are very difficult to be deployed on a single computing machine. In this thesis, we designed a data-intensive computing platform for bioinformatics applications using virtualization technologies and high performance computing (HPC) infrastructures with the concept of multi-tier architecture, which can seamlessly integrate the web user interface (presentation tier), scientific workflow (logic tier) and computing infrastructure (data/computing tier). We demonstrated our platform on two bioinformatics projects. First, we redesigned and deployed the cotton marker database (CMD) (http://www.cottonmarker.org), a centralized web portal in the cotton research community, using the …

Multivalued Subsets Under Information Theory, Indraneel Dabhade

All Theses

In the fields of finance, engineering and varied sciences, Data Mining/ Machine Learning has held an eminent position in predictive analysis. Complex algorithms and adaptive decision models have contributed towards streamlining directed research as well as improve on the accuracies in forecasting. Researchers in the fields of mathematics and computer science have made significant contributions towards the development of this field. Classification based modeling, which holds a significant position amongst the different rule-based algorithms, is one of the most widely used decision making tools. The decision tree has a place of profound significance in classification-based modeling. A number of heuristics …

Accuracy And Multi-Core Performance Of Machine Learning Algorithms For Handwritten Character Recognition, Sumod Mohan

All Theses

There have been considerable developments in the quest for intelligent machines since the beginning of the cybernetics revolution and the advent of computers. In the last two decades with the onset of the internet the developments have been extensive. This quest for building intelligent machines have led into research on the working of human brain, which has in turn led to the development of pattern recognition models which take inspiration in their structure and performance from biological neural networks. Research in creating intelligent systems poses two main problems. The first one is to develop algorithms which can generalize and predict …