Computer Engineering Commons^™

Evaluation Of Robust Deep Learning Pipelines Targeting Low Swap Edge Deployment, David Carter Cornett

Masters Theses

The deep learning technique of convolutional neural networks (CNNs) has greatly advanced the state-of-the-art for computer vision tasks such as image classification and object detection. These solutions rely on large systems leveraging wattage-hungry GPUs to provide the computational power to achieve such performance. However, the size, weight and power (SWaP) requirements of these conventional GPU-based deep learning systems are not suitable when a solution requires deployment to so called "Edge" environments such as autonomous vehicles, unmanned aerial vehicles (UAVs) and smart security cameras.

The objective of this work is to benchmark FPGA-based alternatives to conventional GPU systems that have the …

Action : Adaptive Cache Block Migration In Distributed Cache Architectures, Chandra Sekhar Mummidi

Masters Theses

Increasing number of cores in chip multiprocessors (CMP) result in increasing traffic to last-level cache (LLC). Without commensurate increase in LLC bandwidth, such traffic cannot be sustained resulting in loss of performance. Further, as the number of cores increases, it is necessary to scale up the LLC size; otherwise, the LLC miss rate will rise, resulting in a loss of performance. Unfortunately, for a unified LLC with uniform cache access time, access latency increases with cache size, resulting in performance loss. Previously, researchers have proposed partitioning the cache into multiple smaller caches interconnected by a communication network which increases aggregate …

Benchmarking Small-Dataset Structure-Activity-Relationship Models For Prediction Of Wnt Signaling Inhibition, Mahtab Kokabi

Masters Theses

Quantitative structure-activity relationship (QSAR) models based on machine learning algorithms are powerful tools to expedite drug discovery processes and therapeutics development. Given the cost in acquiring large-sized training datasets, it is useful to examine if QSAR analysis can reasonably predict drug activity with only a small-sized dataset (size < 100) and benchmark these small-dataset QSAR models in application-specific studies. To this end, here we present a systematic benchmarking study on small-dataset QSAR models built for prediction of effective Wnt signaling inhibitors, which are essential to therapeutics development in prevalent human diseases (e.g., cancer). Specifically, we examined a total of 72 two-dimensional (2D) QSAR models based on 4 best-performing algorithms, 6 commonly used molecular fingerprints, and 3 typical fingerprint lengths. We trained these models using a training dataset (56 compounds), benchmarked their performance on 4 figures-of-merit (FOMs), and examined their prediction accuracy using an external validation dataset (14 compounds). Our data show that the model performance is maximized when: 1) molecular fingerprints are selected to provide sufficient, unique, and not overly detailed representations of the chemical structures of drug compounds; 2) algorithms are selected to reduce the number of false predictions due to class imbalance in the dataset; and 3) models are selected to reach balanced performance on all 4 FOMs. These results may provide general guidelines in developing high-performance small-dataset QSAR models for drug activity prediction.

Internet Infrastructures For Large Scale Emulation With Efficient Hw/Sw Co-Design, Aiden K. Gula

Masters Theses

Connected systems are becoming more ingrained in our daily lives with the advent of cloud computing, the Internet of Things (IoT), and artificial intelligence. As technology progresses, we expect the number of networked systems to rise along with their complexity. As these systems become abstruse, it becomes paramount to understand their interactions and nuances. In particular, Mobile Ad hoc Networks (MANET) and swarm communication systems exhibit added complexity due to a multitude of environmental and physical conditions. Testing these types of systems is challenging and incurs high engineering and deployment costs. In this work, we propose a scalable MANET emulation …

A Cloud Infrastructure For Large Scale Health Monitoring In Older Adult Care Facilities, Uchechukwu Gabriel David

Masters Theses

Technology development in the sub-field of older adult care has always been on the back-burner compared to other healthcare areas. But with increasing life expectancy, this is poised to change. With the increasing older adult population, the current older adult care facilities and personnel are struggling to keep up with demand. Research conducted in the Netherlands [1] found 33,000 older adults were awaiting admission into a home for the elderly showing that demand far exceeds availability. This huge demand for older adult care has resulted in a decrease in the quality of care being provided. A recent study involving older …

Design And Simulation Of A Supervisory Control System For Hybrid Manufacturing, Michael Buckley

Masters Theses

The research teams of Dr. Bill Hamel, Dr. Bradley Jared and Dr. Tony Schmitz were tasked by the Office of Naval Research to create a hybrid manufacturing process for a reduced scale model of a naval ship propeller. The base structure of the propeller is created using Wire Arc Additive Manufacturing (WAAM), which is then scanned to compare created geometry to desired geometry. The propeller is then machined down to match the desired geometry. This process is iterated upon until the final product meets design tolerances. Due to the complex nature and numerous industrial machines used in the process, it …

Hardware Acceleration In Image Stitching: Gpu Vs Fpga, Joshua David Edgcombe

Masters Theses

Image stitching is a process where two or more images with an overlapping field of view are combined. This process is commonly used to increase the field of view or image quality of a system. While this process is not particularly difficult for modern personal computers, hardware acceleration is often required to achieve real-time performance in low-power image stitching solutions. In this thesis, two separate hardware accelerated image stitching solutions are developed and compared. One solution is accelerated using a Xilinx Zynq UltraScale+ ZU3EG FPGA and the other solution is accelerated using an Nvidia RTX 2070 Super GPU. The image …

Lecture Video Transformation Through An Intelligent Analysis And Post-Processing System, Xi Wang

Masters Theses

Lecture videos are good sources for people to learn new things. Students commonly use online videos to explore various domains. However, some recorded videos are posted on online platforms without being post-processed due to technology and resource limitations. In this work, we focus on the research of developing an intelligent system to automatically extract essential information, including the main instructor and screen, in a lecture video in several scenarios by using modern deep learning techniques. This thesis aims to combine the extracted essential information to render the videos and generate a new layout with a smaller file size than the …

A Secure Architecture For Defense Against Return Address Corruption, Grayson J. Bruner

Masters Theses

The advent of the Internet of Things has brought about a staggering level of inter-connectivity between common devices used every day. Unfortunately, security is not a high priority for developers designing these IoT devices. Often times the trade-off of security comes at too high of a cost in other areas, such as performance or power consumption. This is especially prevalent in resource-constrained devices, which make up a large number of IoT devices. However, a lack of security could lead to a cascade of security breaches rippling through connected devices. One of the most common attacks used by hackers is return …

Analysis Of Hardware Accelerated Deep Learning And The Effects Of Degradation On Performance, Samuel C. Leach

Masters Theses

As convolutional neural networks become more prevalent in research and real world applications, the need for them to be faster and more robust will be a constant battle. This thesis investigates the effect of degradation being introduced to an image prior to object recognition with a convolutional neural network. As well as experimenting with methods to reduce the degradation and improve performance. Gaussian smoothing and additive Gaussian noise are both analyzed degradation models within this thesis and are reduced with Gaussian and Butterworth masks using unsharp masking and smoothing, respectively. The results show that each degradation is disruptive to the …

Ticknet: A Lightweight Deep Classifier For Tick Recognition, Li Wang

Masters Theses

The world is increasingly controlled by machine learning and deep learning. Deep neural networks are becoming powerful, encroaching on many tasks in computer vision system areas previously seen as the unique domain of humans, such as image classification, object detection, semantic segmentation, and instance segmentation. The success of a deep learning model at a specific application is determined by a sequence of choices, like what kind of deep neural network will be used, what data to be fed into the deep model, and what manners will be adopted to train a deep model.

The goal of this work is to …

Neural Network Supervised And Reinforcement Learning For Neurological, Diagnostic, And Modeling Problems, Donald Wunsch Iii

Masters Theses

“As the medical world becomes increasingly intertwined with the tech sphere, machine learning on medical datasets and mathematical models becomes an attractive application. This research looks at the predictive capabilities of neural networks and other machine learning algorithms, and assesses the validity of several feature selection strategies to reduce the negative effects of high dataset dimensionality. Our results indicate that several feature selection methods can maintain high validation and test accuracy on classification tasks, with neural networks performing best, for both single class and multi-class classification applications. This research also evaluates a proof-of-concept application of a deep-Q-learning network (DQN) to …

Topological Biclustering Artmap, Raghu Yelugam

Masters Theses

”Detection of gene mutations is central for assessing genetic factors affecting disease predisposition, genetic causes of a particular disease, and gene-targeted treatment. DNA microarray methods are widely used to detect mutations by contrasting the expression levels of thousands of genes together under varying experimental conditions. The experimental conditions could be diseased cell states compared with the normal cell states. Biclustering, a robust exploratory data analysis tool, can be applied to microarray data to detect subsets of genes that co-express highly only for a subset of experimental conditions. Such detection is crucial for gaining insights into gene regulatory networks, differential gene …