Physical Sciences and Mathematics Commons^™

Action Recognition, Temporal Localization And Detection In Trimmed And Untrimmed Video, Rui Hou

Electronic Theses and Dissertations

Automatic understanding of videos is one of the most active areas of computer vision research. It has applications in video surveillance, human computer interaction, video sports analysis, virtual and augmented reality, video retrieval etc. In this dissertation, we address four important tasks in video understanding, namely action recognition, temporal action localization, spatial-temporal action detection and video object/action segmentation. This dissertation makes contributions to above tasks by proposing. First, for video action recognition, we propose a category level feature learning method. Our proposed method automatically identifies such pairs of categories using a criterion of mutual pairwise proximity in the (kernelized) feature …

Task Focused Robotic Imitation Learning, Pooya Abolghasemi

Electronic Theses and Dissertations

For many years, successful applications of robotics were the domain of controlled environments, such as industrial assembly lines. Such environments are custom designed for the convenience of the robot and separated from human operators. In recent years, advances in artificial intelligence, in particular, deep learning and computer vision, allowed researchers to successfully demonstrate robots that operate in unstructured environments and directly interact with humans. One of the major applications of such robots is in assistive robotics. For instance, a wheelchair mounted robotic arm can help disabled users in the performance of activities of daily living (ADLs) such as feeding and …

Multi-Touch Detection And Semantic Response On Non-Parametric Rear-Projection Surfaces, Jason Hochreiter

Electronic Theses and Dissertations

The ability of human beings to physically touch our surroundings has had a profound impact on our daily lives. Young children learn to explore their world by touch; likewise, many simulation and training applications benefit from natural touch interactivity. As a result, modern interfaces supporting touch input are ubiquitous. Typically, such interfaces are implemented on integrated touch-display surfaces with simple geometry that can be mathematically parameterized, such as planar surfaces and spheres; for more complicated non-parametric surfaces, such parameterizations are not available. In this dissertation, we introduce a method for generalizable optical multi-touch detection and semantic response on uninstrumented non-parametric …

A Policy Mechanism For Federal Recommendation Of Security Standards For Mobile Devices That Conduct Transactions, Ariel Huckabay

Electronic Theses and Dissertations

The proliferation of mobile devices in the BRIC countries has prompted them to develop policies to manage the security of these devices. In China, mobile devices are a primary tool for payments. As a result, China instituted in 2017 a cyber security policy that applies to mobile devices giving China broad authority to manage cyber threats. The United States has a similar need for a cyber policy. Mobile devices are likely to become a primary payment tool in the United States soon. DHS has also identified a need for more effective security policy in mobile devices for government operations. This …

Probabilistic Record Linkage With Elliptic Curve Operations, Shreya Dhiren Patel

Electronic Theses and Dissertations

Federated query processing for an electronic health record infrastructure enables large epidemiology studies using data integrated from geographically dispersed medical institutions. However, government imposed privacy regulations prohibit disclosure of patient's health record outside the context of clinical care, thereby making it difficult to determine which records correspond to the same entity in the process of query aggregation.

Privacy-preserving record linkage is an actively pursued research area to facilitate the linkage of database records under the constraints of regulations that do not allow the linkage agents to learn sensitive identities of record owners. In earlier works, scalability has been shown to …

Evaluating Software Testing Techniques: A Systematic Mapping Study, Mitchell Mayeda

Electronic Theses and Dissertations

Software testing techniques are crucial for detecting faults in software and reducing the risk of using it. As such, it is important that we have a good understanding of how to evaluate these techniques for their efficiency, scalability, applicability, and effectiveness at finding faults. This thesis enhances our understanding of testing technique evaluations by providing an overview of the state of the art in research. To accomplish this we utilize a systematic mapping study; structuring the field and identifying research gaps and publication trends. We then present a small case study demonstrating how our mapping study can be used to …

Applied Machine Learning For Classification Of Musculoskeletal Inference Using Neural Networks And Component Analysis, Shaswat Sharma

Electronic Theses and Dissertations

Artificial Intelligence (AI) is acquiring more recognition than ever by researchers and machine learning practitioners. AI has found significance in many applications like biomedical research for cancer diagnosis using image analysis, pharmaceutical research, and, diagnosis and prognosis of diseases based on knowledge about patients' previous conditions. Due to the increased computational power of modern computers implementing AI, there has been an increase in the feasibility of performing more complex research.

Within the field of orthopedic biomechanics, this research considers complex time-series dataset of the "sit-to-stand" motion of 48 Total Hip Arthroplasty (THA) patients that was collected by the Human Dynamics …

Application Of Retrograde Analysis To Fighting Games, Kristen Yu

Electronic Theses and Dissertations

With the advent of the fighting game AI competition, there has been recent interest in two-player fighting games. Monte-Carlo Tree-Search approaches currently dominate the competition, but it is unclear if this is the best approach for all fighting games. In this thesis we study the design of two-player fighting games and the consequences of the game design on the types of AI that should be used for playing the game, as well as formally define the state space that fighting games are based on. Additionally, we also characterize how AI can solve the game given a simultaneous action game model, …

Building An Automated Q-A System Using Online Forums As Knowledge Bases, Kyle Moore

Electronic Theses and Dissertations

Question-Answer systems traditionally use expensive and difficult to produce structured knowledge bases. Recent systems have used unstructured natural language sources as their datasets, but most of those sources have been overly broad or difficult to extend. Online forums are a largely untapped source of information that can provide both depth and breadth when limited to a specific domain, as well as being adaptive to the introduction of new information. In this paper, I conjecture that online forums can be similarly and effectively used as an unstructured knowledge base for Question-Answer systems. I use a relatively simple summarization-based approach to analyze …

Utilizing Edge In Iot And Video Streaming Applications To Reduce Bottlenecks In Internet Traffic, Kutalmis Akpinar

Electronic Theses and Dissertations

There is a large increase in the surge of data over Internet due to the increasing demand on multimedia content. It is estimated that 80% of Internet traffic will be video by 2022, according to a recent study. At the same time, IoT devices on Internet will double the human population. While infrastructure standards on IoT are still nonexistent, enterprise solutions tend to encourage cloud-based solutions, causing an additional surge of data over the Internet. This study proposes solutions to bring video traffic and IoT computation back to the edges of the network, so that costly Internet infrastructure upgrades are …

Learning Internal State Memory Representations From Observation, Josiah Wong

Electronic Theses and Dissertations

Learning from Observation (LfO) is a machine learning paradigm that mimics how people learn in daily life: learning how to do something simply by watching someone else do it. LfO has been used in various applications, from video game agent creation to driving a car, but it has always been limited by the inability of an observer to know what a performing entity chooses to remember as they act in an environment. Various methods have either ignored the effects of memory or otherwise made simplistic assumptions about its structure. In this dissertation, we propose a new method, Memory Composition Learning, …

Scalable Network Design And Management With Decentralized Software-Defined Networking, Kuldip Singh Atwal

Electronic Theses and Dissertations

Network softwarization is among the most significant innovations of computer networks in the last few decades. The lack of uniform and programmable interfaces for network management led to the design of OpenFlow protocol for the university campuses and enterprise networks. This breakthrough coupled with other similar efforts led to an emergence of two complementary but independent paradigms called software-defined networking (SDN) and network function virtualization (NFV). As of this writing, these paradigms are becoming the de-facto norms of wired and wireless networks alike. This dissertation mainly addresses the scalability aspect of SDN for multiple network types. Although centralized control and …

Approximate In-Memory Computing On Rerams, Salman Anwar Khokhar

Electronic Theses and Dissertations

Computing systems have seen tremendous growth over the past few decades in their capabilities, efficiency, and deployment use cases. This growth has been driven by progress in lithography techniques, improvement in synthesis tools, architectures and power management. However, there is a growing disparity between computing power and the demands on modern computing systems. The standard Von-Neuman architecture has separate data storage and data processing locations. Therefore, it suffers from a memory-processor communication bottleneck, which is commonly referred to as the 'memory wall'. The relatively slower progress in memory technology compared with processing units has continued to exacerbate the memory wall …

Training Neural Networks Through The Integration Of Evolution And Gradient Descent, Gregory Morse

Electronic Theses and Dissertations

Neural networks have achieved widespread adoption due to both their applicability to a wide range of problems and their success relative to other machine learning algorithms. The training of neural networks is achieved through any of several paradigms, most prominently gradient-based approaches (including deep learning), but also through up-and-coming approaches like neuroevolution. However, while both of these neural network training paradigms have seen major improvements over the past decade, little work has been invested in developing algorithms that incorporate the advances from both deep learning and neuroevolution. This dissertation introduces two new algorithms that are steps towards the integration of …

Optimization Algorithms For Deep Learning Based Medical Image Segmentations, Aliasghar Mortazi

Electronic Theses and Dissertations

Medical image segmentation is one of the fundamental processes to understand and assess the functionality of different organs and tissues as well as quantifying diseases and helping treatment planning. With ever increasing number of medical scans, the automated, accurate, and efficient medical image segmentation is as unmet need for improving healthcare. Recently, deep learning has emerged as one the most powerful methods for almost all image analysis tasks such as segmentation, detection, and classification and so in medical imaging. In this regard, this dissertation introduces new algorithms to perform medical image segmentation for different (a) imaging modalities, (b) number of …

Multi-Modal Interfaces For Sensemaking Of Graph-Connected Datasets, Anthony Wehrer

Electronic Theses and Dissertations

The visualization of hypothesized evolutionary processes is often shown through phylogenetic trees. Given evolutionary data presented in one of several widely accepted formats, software exists to render these data into a tree diagram. However, software packages commonly in use by biologists today often do not provide means to dynamically adjust and customize these diagrams for studying new hypothetical relationships, and for illustration and publication purposes. Even where these options are available, there can be a lack of intuitiveness and ease-of-use. The goal of our research is, thus, to investigate more natural and effective means of sensemaking of the data with …

Visual-Textual Video Synopsis Generation, Aidean Sharghi Karganroodi

Electronic Theses and Dissertations

In this dissertation we tackle the problem of automatic video summarization. Automatic summarization techniques enable faster browsing and indexing of large video databases. However, due to the inherent subjectivity of the task, no single video summarizer fits all users unless it adapts to individual user's needs. To address this issue, we introduce a fresh view on the task called "Query-focused'' extractive video summarization. We develop a supervised model that takes as input a video and user's preference in form of a query, and creates a summary video by selecting key shots from the original video. We model the problem as …

A Study Of Perceptions On Incident Response Exercises, Information Sharing, Situational Awareness, And Incident Response Planning In Power Grid Utilities, Joseph Garmon

Electronic Theses and Dissertations

The power grid is facing increasing risks from a cybersecurity attack. Attacks that shut off electricity in Ukraine have already occurred, and successful compromises of the power grid that did not shut off electricity to customers have been privately disclosed in North America. The objective of this study is to identify how perceptions of various factors emphasized in the electric sector affect incident response planning. Methods used include a survey of 229 power grid personnel and the use of partial least squares structural equation modeling to identify causal relationships. This study reveals the relationships between perceptions by personnel responsible for …

Blockchain-Driven Secure And Transparent Audit Logs, Ashar Ahmad

Electronic Theses and Dissertations

In enterprise business applications, large volumes of data are generated daily, encoding business logic and transactions. Those applications are governed by various compliance requirements, making it essential to provide audit logs to store, track, and attribute data changes. In traditional audit log systems, logs are collected and stored in a centralized medium, making them prone to various forms of attacks and manipulations, including physical access and remote vulnerability exploitation attacks, and eventually allowing for unauthorized data modification, threatening the guarantees of audit logs. Moreover, such systems, and given their centralized nature, are characterized by a single point of failure. To …

Parameter Estimation Of Stochastic Models Against Probabilistic Temporal Logic Behavioral Specifications, Arfeen Khalid

Electronic Theses and Dissertations

The inherent behavioral variability exhibited by stochastic systems makes it a challenging task for human experts to manually analyze them. Computational modeling of such systems helps in investigating and predicting the behaviors of their underlying processes but at the same time introduces the presence of several unknown parameters. A key challenge faced in this scenario is to determine the values of these unknown parameters against known behavioral specifications. The solutions that have been presented so far estimate the parameters of a given model against a single specification whereas a correct model is expected to satisfy all the behavioral specifications when …

Collaborative Artificial Intelligence Algorithms For Medical Imaging Applications, Naji Khosravan

Electronic Theses and Dissertations

In this dissertation, we propose novel machine learning algorithms for high-risk medical imaging applications. Specifically, we tackle current challenges in radiology screening process and introduce cutting-edge methods for image-based diagnosis, detection and segmentation. We incorporate expert knowledge through eye-tracking, making the whole process human-centered. This dissertation contributes to machine learning, computer vision, and medical imaging research by: 1) introducing a mathematical formulation of radiologists level of attention, and sparsifying their gaze data for a better extraction and comparison of search patterns. 2) proposing novel, local and global, image analysis algorithms. Imaging based diagnosis and pattern analysis are "high-risk" Artificial Intelligence …

Reducing The Large Class Code Smell By Applying Design Patterns, Bayan Turkistani

Electronic Theses and Dissertations

Software systems need continuous developing to cope and keep up with everchanging requirements. Source code quality affects the software development costs. In software refactoring object-oriented systems, Large Class, in particular, hinder the maintenance of a system by letting it difficult for software developers to understand and perform modifications. Also, it is making the development process labor-intensive and time-wasting. Reducing the Large Class code smell by applying design patterns can make the refactoring process more manageable, ease developing the system and decrease the effort required for the maintaining of software. To guarantee object-oriented software stays clear to read, understand and modify …

Instance Segmentation And Object Detection In Road Scenes Using Inverse Perspective Mapping Of 3d Point Clouds And 2d Images, Chungyup Lee

Electronic Theses and Dissertations

The instance segmentation and object detection are important tasks in smart car applications. Recently, a variety of neural network-based approaches have been proposed. One of the challenges is that there are various scales of objects in a scene, and it requires the neural network to have a large receptive field to deal with the scale variations. In other words, the neural network must have deep architectures which slow down computation. In smart car applications, the accuracy of detection and segmentation of vehicle and pedestrian is hugely critical. Besides, 2D images do not have distance information but enough visual appearance. On …

Towards Misleading Connection Mining, Md Main Uddin Rony

Electronic Theses and Dissertations

This study introduces a new Natural Language Generation (NLG) task – Unit Claim Identification. The task aims to extract every piece of verifiable information from a headline. The Unit Claim identification has applications in other domains; such as fact-checking where the identification of each verifiable information from a check-worthy statement can lead to an effective fact-check. Moreover, the extracting of the unit claims from headlines can identify a misleading news article, by mapping evidence from contents. For addressing the unit claim identification problem, we outlined a set of guidelines for data annotation, arranged in-house training for the annotators and obtained …

Optimizing The Performance Of Complex Engineering Systems Aided By Artificial Neural Networks, Khalil Qatu

Electronic Theses and Dissertations

In the first problem Polyetherimide graphene nanoplatelets papers (PEIGNP) were tested with different graphene loadings varying from 0-97 weight percent (WT%). The resulting stress-strain curves were utilized to develop two ANN models. Stress-controlled and strain-controlled models. Both models shoan excellent correlation to the experimental. Several Mechanical properties were calculated from the predicted stress-strain curves namely; toughness maximum strength maximum strain and maximum tangent modulus. Both models captured the same overall behavior of the PEIGNP composite. However the strain-controlled model was found to predict lower stress than the stress-controlled model. Finally a Graphical User Interface (GUI) was developed to aid in …

Performance Evaluation Of Blocking And Non-Blocking Concurrent Queues On Gpus, Hossein Pourmeidani

Electronic Theses and Dissertations

The efficiency of concurrent data structures is crucial to the performance of multi-threaded programs in shared-memory systems. The arbitrary execution of concurrent threads, however, can result in an incorrect behavior of these data structures. Graphics Processing Units (GPUs) have appeared as a powerful platform for high-performance computing. As regular data-parallel computations are straightforward to implement on traditional CPU architectures, it is challenging to implement them in a SIMD environment in the presence of thousands of active threads on GPU architectures. In this thesis, we implement a concurrent queue data structure and evaluate its performance on GPUs to understand how it …

An Approach To Semi-Autonomous Indoor Drone System: Software Architecture And Integration Testing, Shobhan Singh

Electronic Theses and Dissertations

To address these problems, we establish a semi-autonomous functionality by removing the RC transmitter, and remotely connecting the Drone System to track status and executing user-based input commands. In order to resolve the limitation in hardware connections on the Flight Controller, we integrated the sonar sensor into a companion computer, from where the data is continuously fed to an embedded system through MAVLink (Micro Aerial Vehicle Link) network communication protocol. In this study, we also implemented a modular architecture which enables scalable integration of sensor modules into the Drone System to streamline the process of development, deployment, testing and debugging.

Utilizing Various Neural Network Architectures To Play A Game Developed For Human Players, Michael Blake Arender

Electronic Theses and Dissertations

Neural Networks have received an explosive amount of attention and interest in recent years. Despite the fact that Neural Network algorithms having existed for many decades, it was not until recent advances in computer hardware that they saw widespread use. This is in no small part due to the success these algorithms have had in tasks ranging from image classification, voice recognition, game theory, and many other applications. Thanks to recent strides in hardware development, most importantly in the advancements in Graphics Processor Units including the capabilities of modern GPU Computing, Neural Networks are now capable of solving tasks at …

Scheduling Irregular Workloads On Gpus, David Arthur Troendle

Electronic Theses and Dissertations

This doctoral research aims at understanding the nature of the overhead for data irregular GPU workloads, proposing a solution, and examining the consequences of the result. We propose a novel, retry-free GPU workload scheduler for irregular workloads. When used in a Breadth First Search (BFS) algorithm, the proposed simple, monolithic concurrent queue scales to within 10% of ideal scalability on AMD’s Fiji GPU with 14,336 active threads. The dissertation presents an important finding that the retry overhead associated with Compare and Swap (CAS) operations is the principle reason why concurrent queues do not scale well as the number of clients …

Improving Random Forests By Feature Dependence Analysis, Silu Zhang

Electronic Theses and Dissertations

Random forests (RFs) have been widely used for supervised learning tasks because of their high prediction accuracy good model interpretability and fast training process. However they are not able to learn from local structures as convolutional neural networks (CNNs) do when there exists high dependency among features. They also cannot utilize features that are jointly dependent on the label but marginally independent of it. In this dissertation we present two approaches to address these two problems respectively by dependence analysis. First a local feature sampling (LFS) approach is proposed to learn and use the locality information of features to group …