Engineering Commons^™

Assessing High Dynamic Range Imagery Performance For Object Detection In Maritime Environments, Erasmo Landaeta

Doctoral Dissertations and Master's Theses

The field of autonomous robotics has benefited from the implementation of convolutional neural networks in vision-based situational awareness. These strategies help identify surface obstacles and nearby vessels. This study proposes the introduction of high dynamic range cameras on autonomous surface vessels because these cameras capture images at different levels of exposure revealing more detail than fixed exposure cameras. To see if this introduction will be beneficial for autonomous vessels this research will create a dataset of labeled high dynamic range images and single exposure images, then train object detection networks with these datasets to compare the performance of these networks. …

Adversarial Training Of Deep Neural Networks, Anabetsy Termini

CCE Theses and Dissertations

Deep neural networks used for image classification are highly susceptible to adversarial attacks. The de facto method to increase adversarial robustness is to train neural networks with a mixture of adversarial images and unperturbed images. However, this method leads to robust overfitting, where the network primarily learns to recognize one specific type of attack used to generate the images while remaining vulnerable to others after training. In this dissertation, we performed a rigorous study to understand whether combinations of state of the art data augmentation methods with Stochastic Weight Averaging improve adversarial robustness and diminish adversarial overfitting across a wide …

A Deep Reinforcement Learning Approach With Prioritized Experience Replay And Importance Factor For Makespan Minimization In Manufacturing, Jose Napoleon Martinez

LSU Doctoral Dissertations

In this research, we investigated the application of deep reinforcement learning (DRL) to a common manufacturing scheduling optimization problem, max makespan minimization. In this application, tasks are scheduled to undergo processing in identical processing units (for instance, identical machines, machining centers, or cells). The optimization goal is to assign the jobs to be scheduled to units to minimize the maximum processing time (i.e., makespan) on any unit.

Machine learning methods have the potential to "learn" structures in the distribution of job times that could lead to improved optimization performance and time over traditional optimization methods, as well as to adapt …

Evaluating Neural Network Decoder Performance For Quantum Error Correction Using Various Data Generation Models, Brett M. Martin

Theses and Dissertations

Neural networks have been shown in the past to perform quantum error correction (QEC) decoding with greater accuracy and efficiency than algorithmic decoders. Because the qubits in a quantum computer are volatile and only usable on the order of milliseconds before they decohere, a means of fast quantum error correction is necessary in order to correct data qubit errors within the time budget of a quantum algorithm. Algorithmic decoders are good at resolving errors on logical qubits with only a few data qubits, but are less efficient in systems containing more data qubits. With neural network decoders, practical quantum computation …

Extractive Text Summarization On Single Documents Using Deep Learning, Shehab Mostafa Abdel-Salam Mohamed

Theses and Dissertations

The task of summarization can be categorized into two methods, extractive and abstractive summarization. Extractive approach selects highly meaningful sentences to form a summary while the abstractive approach interprets the original document and generates the summary in its own words. The task of generating a summary, whether extractive or abstractive, has been studied with different approaches such as statistical-based, graph-based, and deep-learning based approaches. Deep learning has achieved promising performance in comparison with the classical approaches and with the evolution of neural networks such as the attention network or commonly known as the Transformer architecture, there are potential areas for …

Deep Learning Based Speech Enhancement And Its Application To Speech Recognition, Ju Lin

All Dissertations

Speech enhancement is the task that aims to improve the quality and the intelligibility of a speech signal that is degraded by ambient noise and room reverberation. Speech enhancement algorithms are used extensively in many audio- and communication systems, including mobile handsets, speech recognition, speaker verification systems and hearing aids. Recently, deep learning has achieved great success in many applications, such as computer vision, nature language processing and speech recognition. Speech enhancement methods have been introduced that use deep-learning techniques, as these techniques are capable of learning complex hierarchical functions using large-scale training data. This dissertation investigates the deep learning …

Predicting Imports In Java Code With Graph Neural Networks, Aleksandr Fedchin

Senior Projects Spring 2020

Programmers tend to split their code into multiple files or sub-modules. When a program is executed, these sub-modules interact to produce the desired effect. One can, therefore, represent programs with graphs, where each node corresponds to some file and each edge corresponds to some relationship between files, such as two files being located in the same package or one file importing the content of another. This project trains Graph Neural Networks on such graphs to learn to predict future imports in Java programs and shows that Graph Neural Networks outperform various baseline methods by a wide margin.

Design Of A Novel Wearable Ultrasound Vest For Autonomous Monitoring Of The Heart Using Machine Learning, Garrett G. Goodman

Browse all Theses and Dissertations

As the population of older individuals increases worldwide, the number of people with cardiovascular issues and diseases is also increasing. The rate at which individuals in the United States of America and worldwide that succumb to Cardiovascular Disease (CVD) is rising as well. Approximately 2,303 Americans die to some form of CVD per day according to the American Heart Association. Furthermore, the Center for Disease Control and Prevention states that 647,000 Americans die yearly due to some form of CVD, which equates to one person every 37 seconds. Finally, the World Health Organization reports that the number one cause of …

Multiple Face Detection And Recognition System Design Applying Deep Learning In Web Browsers Using Javascript, Cristhian Gabriel Espinosa Sandoval

Computer Science and Computer Engineering Undergraduate Honors Theses

Deep learning has advanced progressively in the last years and now demonstrates state-of-the-art performance in various fields. In the era of big data, transformation of data into valuable knowledge has become one of the most important challenges in computing. Therefore, we will review multiple algorithms for face recognition that have been researched for a long time and are maturely developed, and analyze deep learning, presenting examples of current research.

To provide a useful and comprehensive perspective, in this paper we categorize research by deep learning architecture, including neural networks, convolutional neural networks, depthwise Separable Convolutions, densely connected convolutional networks, and …

Improved Study Of Side-Channel Attacks Using Recurrent Neural Networks, Muhammad Abu Naser Rony Chowdhury

Boise State University Theses and Dissertations

Differential power analysis attacks are special kinds of side-channel attacks where power traces are considered as the side-channel information to launch the attack. These attacks are threatening and significant security issues for modern cryptographic devices such as smart cards, and Point of Sale (POS) machine; because after careful analysis of the power traces, the attacker can break any secured encryption algorithm and can steal sensitive information.

In our work, we study differential power analysis attack using two popular neural networks: Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN). Our work seeks to answer three research questions(RQs):

RQ1: Is it …

Investigating Semantic Properties Of Images Generated From Natural Language Using Neural Networks, Samuel Ward Schrader

Boise State University Theses and Dissertations

This work explores the attributes, properties, and potential uses of generative neural networks within the realm of encoding semantics. It works toward answering the questions of: If one uses generative neural networks to create a picture based on natural language, does the resultant picture encode the text's semantics in a way a computer system can process? Could such a system be more precise than current solutions at detecting, measuring, or comparing semantic properties of generated images, and thus their source text, or their source semantics?

This work is undertaken in the hope that detecting previously unknown properties, or better understanding …

Hyper-Parameter Optimization Of A Convolutional Neural Network, Steven H. Chon

Theses and Dissertations

In the world of machine learning, neural networks have become a powerful pattern recognition technique that gives a user the ability to interpret high-dimensional data whereas conventional methods, such as logistic regression, would fail. There exists many different types of neural networks, each containing its own set of hyper-parameters that are dependent on the type of analysis required, but the focus of this paper will be on the hyper-parameters of convolutional neural networks. Convolutional neural networks are commonly used for classifications of visual imagery. For example, if you were to build a network for the purpose of predicting a specific …

Cyber Data Anomaly Detection Using Autoencoder Neural Networks, Spencer A. Butt

Theses and Dissertations

The Department of Defense requires a secure presence in the cyber domain to successfully execute its stated mission of deterring war and protecting the security of the United States. With potentially millions of logged network events occurring on defended networks daily, a limited staff of cyber analysts require the capability to identify novel network actions for security adjudication. The detection methodology proposed uses an autoencoder neural network optimized via design of experiments for the identification of anomalous network events. Once trained, each logged network event is analyzed by the neural network and assigned an outlier score. The network events with …

Life Long Learning In Sparse Learning Environments, John Reeder

Electronic Theses and Dissertations

Life long learning is a machine learning technique that deals with learning sequential tasks over time. It seeks to transfer knowledge from previous learning tasks to new learning tasks in order to increase generalization performance and learning speed. Real-time learning environments in which many agents are participating may provide learning opportunities but they are spread out in time and space outside of the geographical scope of a single learning agent. This research seeks to provide an algorithm and framework for life long learning among a network of agents in a sparse real-time learning environment. This work will utilize the robust …

Solving The Vehicle Re-Identification Problem By Using Neural Networks, Tanweer Rashid

Computational Modeling & Simulation Engineering Theses & Dissertations

Vehicle re-identification is the process by which vehicle attributes measured at one point on a road network are compared to vehicle attributes measured at another point in an effort to match vehicles without using any unique identifiers such as license plate numbers. A match is made if the two measurements are estimated to belong to the same vehicle. Vehicle attributes can be sensor readings such as loop induction signatures, or they can also be actual vehicle characteristics such as length, weight, number of axles, etc. This research makes use of vehicle length, travel time, axle spacing and axle weights for …

Architecture For Intelligent Power Systems Management, Optimization, And Storage., J. Chris Foreman

Electronic Theses and Dissertations

The management of power and the optimization of systems generating and using power are critical technologies. A new architecture is developed to advance the current state of the art by providing an intelligent and autonomous solution for power systems management. The architecture is two-layered and implements a decentralized approach by defining software objects, similar to software agents, which provide for local optimization of power devices such as power generating, storage, and load devices. These software device objects also provide an interface to a higher level of optimization. This higher level of optimization implements the second layer in a centralized approach …

A Neat Approach To Genetic Programming, Adelein Rodriguez

Electronic Theses and Dissertations

The evolution of explicitly represented topologies such as graphs involves devising methods for mutating, comparing and combining structures in meaningful ways and identifying and maintaining the necessary topological diversity. Research has been conducted in the area of the evolution of trees in genetic programming and of neural networks and some of these problems have been addressed independently by the different research communities. In the domain of neural networks, NEAT (Neuroevolution of Augmenting Topologies) has shown to be a successful method for evolving increasingly complex networks. This system's success is based on three interrelated elements: speciation, marking of historical information in …

Development Of Self-Adaptive Back Propagation And Derivative Free Training Algorithms In Artificial Neural Networks, Shamsuddin Ahmed

Theses: Doctorates and Masters

Three new iterative, dynamically self-adaptive, derivative-free and training parameter free artificial neural network (ANN) training algorithms are developed. They are defined as self-adaptive back propagation, multi-directional and restart ANN training algorithms. The descent direction in self-adaptive back propagation training is determined implicitly by a central difference approximation scheme, which chooses its step size according to the convergence behavior of the error function. This approach trains an ANN when the gradient information of the corresponding error function is not readily available. The self- adaptive variable learning rates per epoch are determined dynamically using a constrained interpolation search. As a result, appropriate …