Digital Commons Network^™

Deep Learning Based Sound Event Detection And Classification, Alireza Nasiri

Theses and Dissertations

Hearing sense has an important role in our daily lives. During the recent years, there has been many studies to transfer this capability to the computers. In this dissertation, we design and implement deep learning based algorithms to improve the ability of the computers in recognizing the different sound events.

In the first topic, we investigate sound event detection, which identifies the time boundaries of the sound events in addition to the type of the events. For sound event detection, we propose a new method, AudioMask, to benefit from the object-detection techniques in computer vision. In this method, we convert …

Deep Learning Based Models For Classification From Natural Language Processing To Computer Vision, Xianshan Qu

Theses and Dissertations

With the availability of large scale data sets, researchers in many different areas such as natural language processing, computer vision, recommender systems have started making use of deep learning models and have achieved great progress in recent years. In this dissertation, we study three important classification problems based on deep learning models.

First, with the fast growth of e-commerce, more people choose to purchase products online and browse reviews before making decisions. It is essential to build a model to identify helpful reviews automatically. Our work is inspired by the observation that a customer's expectation of a review can be …

Learning How To Search: Generating Effective Test Cases Through Adaptive Fitness Function Selection, Hussein Khalid Almulla

Theses and Dissertations

Search-based test generation is guided by feedback from one or more fitness functions— scoring functions that judge solution optimality. Choosing informative fitness functions is crucial to meeting the goals of a tester. Unfortunately, many goals—such as forcing the class-under-test to throw exceptions, increasing test suite diversity, and attaining Strong Mutation Coverage—do not have effective fitness function formulations. We propose that meeting such goals requires treating fitness function identification as a secondary optimization step. An adaptive algorithm that can vary the selection of fitness functions could adjust its selection throughout the generation process to maximize goal attainment, based on the current …

Detecting The Intent Of Email Using Embeddings, Deep Learning And Transfer Learning, Zaid Alibadi

Theses and Dissertations

Throughout the years' several strategies and tools were proposed and developed to help the users cope with the problem of email overload, but each of these solutions had its own limitations and, in some cases, contribute to further problems. One major theme that encapsulates many of these solutions is automatically classifying emails into predefined categories (ex: Finance, Sport, Promotion, etc.) then move/tag the incoming email to that particular category. In general, these solutions have two main limitations: 1) they need to adapt to changing user’s behavior. 2) they require handcrafted features engineering which in turn need a lot of time, …

Regularized Deep Network Learning For Multi-Label Visual Recognition, Hao Guo

Theses and Dissertations

This dissertation is focused on the task of multi-label visual recognition, a fundamental task of computer vision. It aims to tell the presence of multiple visual classes from the input image, where the visual classes, such as objects, scenes, attributes, etc., are usually defined as image labels. Due to the prosperous deep networks, this task has been widely studied and significantly improved in recent years. However, it remains a challenging task due to appearance complexity of multiple visual contents co-occurring in one image. This research explores to regularize the deep network learning for multi-label visual recognition.

First, an attention concentration …

The Utility Of Multiple Structure Torsion Angle Alignment In Protein Active Site Description (Asd), Devaun L. Mcfarland

Theses and Dissertations

Proteins are responsible for various functions throughout organisms, or within the systems, they operate. Active-sites or functional/ binding sites are regions responsible for activity in a protein; they serve as a catalyst for reactions, attach or bind to other molecules (ligands), and maintain function. With the profusion of protein sequence and structure data, it's increasingly relevant to develop automated methods of identifying and investigating active-sites for proteins. Active-sites identification will have a direct impact: in better understanding molecular basis for diseases, assisting in drug design, the study of targeting mutants, and for functional annotation of unknown proteins. The proper knowledge …

Providing Predictable Performance During Network Contingencies, Phani Krishna Penumarthi

Theses and Dissertations

In IP backbone networks, packets may get dropped due to: i) lack of viable next hops when a link/router fails, ii) forwarding loops during network convergence, and iii) buffer overflows in case of congestion. Similarly, packets may be lost in wireless networks due to variations in signal strength between a pair of mobile nodes. This dissertation explores the possibility of providing a predictable performance during such network contingencies in wired backbone networks and robotic wireless networks.

First, we study the feasibility of developing a combination of local reroute and global update mechanisms that can achieve loop-free convergence, while performing disruption-free …

Algorithmic Robot Design: Label Maps, Procrustean Graphs, And The Boundary Of Non-Destructiveness, Shervin Ghasemlou

Theses and Dissertations

This dissertation is focused on the problem of algorithmic robot design. The process of designing a robot or a team of robots that can reliably accomplish a task in an environment requires several key elements. How the problem is formulated can play a big role in the design process. The ability of the model to correctly reflect the environment, the events, and different pieces of the problem is crucial. Another key element is the ability of the model to show the relationship between different designs of a single system. These two elements can enable design algorithms to navigate through the …

Smart Sensing Enabled Secure And Usable Pairing And Authentication, Xiaopeng Li

Theses and Dissertations

Internet of Things (IoT) technologies have made our lives more convenient and better informed by sensing and monitoring our surroundings. Security applications, such as device pairing and user authentication, are the fundamentals for building a trustworthy smart environment. A secure and convenient pairing approach is critical to IoT enabled applications, as pairing is to establish a secure wireless communication channel for devices. Besides, a smart environment usually has multiple people (e.g., patients and doctors in a hospital), who have physical access to the deployed IoT devices and sensitive dumb objects (e.g., a cabinet storing medical records); but not all of …

Moving-Camera Video Content Analysis Via Action Recognition And Homography Transformation, Yang Mi

Theses and Dissertations

Moving-camera video content analysis aims at interpreting useful information in videos taken by moving cameras, including wearable cameras and handy cameras. It is an essential problem in computer vision, and plays an important role in many real-life applications, including understanding social difficulties and enhancing public security. In this work, we study three sub-problems of moving-camera video content analysis, including two sub-problems for the analysis on wearable-camera videos which are a special type of moving camera videos: recognizing general actions and recognizing microactions in wearable-camera videos. And, the third sub-problem is estimating homographies along moving-camera videos.

Recognizing general actions in wearable-camera …

From Cellular To Holistic: Development Of Algorithms To Study Human Health And Diseases, Casey Anne Cole

Theses and Dissertations

The development of theoretical computational methods and their application has become widespread in the world today. In this dissertation, I present my work in the creation of models to detect and describe complex biological and health related problems. The first major part of my work centers around the creation and enhancement of methods to calculate protein structure and dynamics. To this end, substantial enhancement has been made to the software package REDCRAFT to better facilitate its usage in protein structure calculation. The enhancements have led to an overall increase in its ability to characterize proteins under difficult conditions such as …

An Overlay Architecture For Pattern Matching, Rasha Elham Karakchi

Theses and Dissertations

Deterministic and Non-deterministic Finite Automata (DFA and NFA) comprise the fundamental unit of work for many emerging big data applications, motivating recent efforts to develop Domain-Specific Architectures (DSAs) to exploit fine-grain parallelism available in automata workloads.

This dissertation presents NAPOLY (Non-Deterministic Automata Processor Over- LaY), an overlay architecture and associated software that attempt to maximally exploit on-chip memory parallelism for NFA evaluation. In order to avoid an upper bound in NFA size that commonly affects prior efforts, NAPOLY is optimized for runtime reconfiguration, allowing for full reconfiguration in 10s of microseconds. NAPOLY is also parameterizable, allowing for offline generation of …

A Machine Learning Based Approach To Accelerate Catalyst Discovery, Asif Jamil Chowdhury

Theses and Dissertations

Computational catalysis, in contrast to experimental catalysis, uses approximations such as density functional theory (DFT) to compute properties of reaction intermediates. But DFT calculations for a large number of surface species on variety of active site models are resource intensive. In this work, we are building a machine learning based predictive framework for adsorption energies of intermediate species, which can reduce the computational overhead significantly. Our work includes the study and development of appropriate machine learning models and effective fingerprints or descriptors to predict energies accurately for different scenarios. Furthermore, Bayesian inverse problem, that integrates experimental catalysis with its computational …

Properties, Learning Algorithms, And Applications Of Chain Graphs And Bayesian Hypergraphs, Mohammad Ali Javidian

Theses and Dissertations

Probabilistic graphical models (PGMs) use graphs, either undirected, directed, or mixed, to represent possible dependencies among the variables of a multivariate probability distri- bution. PGMs, such as Bayesian networks and Markov networks, are now widely accepted as a powerful and mature framework for reasoning and decision making under uncertainty in knowledge-based systems. With the increase of their popularity, the range of graphical models being investigated and used has also expanded. Several types of graphs with dif- ferent conditional independence interpretations - also known as Markov properties - have been proposed and used in graphical models.

The graphical structure of a …

Person Identification With Convolutional Neural Networks, Kang Zheng

Theses and Dissertations

Person identification aims at matching persons across images or videos captured by different cameras, without requiring the presence of persons’ faces. It is an important problem in computer vision community and has many important real-world applica- tions, such as person search, security surveillance, and no-checkout stores. However, this problem is very challenging due to various factors, such as illumination varia- tion, view changes, human pose deformation, and occlusion. Traditional approaches generally focus on hand-crafting features and/or learning distance metrics for match- ing to tackle these challenges. With Convolutional Neural Networks (CNNs), feature extraction and metric learning can be combined in …

Machine Learning Based Ultra High Carbon Steel Image Segmentation, Sumith Kuttiyil Suresh

Theses and Dissertations

Mechanical and structural properties of ultra-high carbon steel are determined by their microstructures composed of constituents such as pearlite and spheroidites. Locating micro constituents and quantitatively measuring its presence is key for material researchers to study the physical properties of the carbon steel materials. This micrograph analysis is currently done manually and subjectively by material scientists, which is tedious and time-consuming. Here we propose to apply the image segmentation algorithm called U-Net to achieve automated labeling of steel microstructures on a subset of ultra- high carbon steel image dataset containing pearlite and spheroidite as the primary micro constituents. Our work …

Improving Person-Independent Facial Expression Recognition Using Deep Learning, Jie Cai

Theses and Dissertations

Over the past few years, deep learning, e.g., Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs), have shown promise on facial expression recog- nition. However, the performance degrades dramatically especially in close-to-real-world settings due to high intra-class variations and high inter-class similarities introduced by subtle facial appearance changes, head pose variations, illumination changes, occlusions, and identity-related attributes, e.g., age, race, and gender. In this work, we developed two novel CNN frameworks and one novel GAN approach to learn discriminative features for facial expression recognition.

First, a novel island loss is proposed to enhance the discriminative power of learned deep …

Insider’S Misuse Detection: From Hidden Markov Model To Deep Learning, Ahmed Saaudi

Theses and Dissertations

Malicious insiders increasingly affect organizations by leaking classified data to unautho- rized entities. Detecting insiders’ misuses in computer systems is a challenging problem. In this dissertation, we propose two approaches to detect such threats: a probabilistic graph- ical model-based approach and a deep learning-based approach. We investigate the logs of computer-based activities to discover patterns of misuse. We model user’s behaviors as sequences of computer-based events.

For our probabilistic graphical model-based approach, we propose an unsupervised model for insider’s misuse detection. That is, we develop Stochastic Gradient Descent method to learn Hidden Markov Models (SGD-HMM) with the goal of analyzing …

Stacked Modelling Framework, Kareem Abdelfatah

Theses and Dissertations

The thesis develops a predictive modeling framework based on stacked Gaussian processes and applies it to two main applications in environmental and chemical en- gineering. First, a network of independently trained Gaussian processes (StackedGP) is introduced to obtain analytical predictions of quantities of interest (model out- puts) with quantified uncertainties. StackedGP framework supports component- based modeling in different fields such as environmental and chemical science, en- hances predictions of quantities of interest through a cascade of intermediate predic- tions usually addressed by cokriging, and propagates uncertainties through emulated dynamical systems driven by uncertain forcing variables. By using analytical first and …

Semantic Segmentation Considering Image Degradation, Global Context, And Data Balancing, Dazhou Guo

Theses and Dissertations

Recently, semantic segmentation – assigning a categorical label to each pixel in an im- age – plays an important role in image understanding applications, e.g., autonomous driving, human-machine interaction and medical imaging. Semantic segmentation has made progress by using the deep convolutional neural networks, which are sur- passing the traditional methods by a large margin. Despite the success of the deep convolutional neural networks (CNNs), there remain three major challenges.

The first challenge is how to segment the degraded images semantically, i.e., de- graded image semantic segmentation. In general, image degradations increase the difficulty of semantic segmentation, usually leading to …

Phylogenetic Reconstruction Analysis On Gene Order And Copy Number Variation, Ruofan Xia

Theses and Dissertations

Genome rearrangement is known as one of the main evolutionary mechanisms on the genomic level. Phylogenetic analysis based on rearrangement played a crucial role in biological research in the past decades, especially with the increasing avail- ability of fully sequenced genomes. In general, phylogenetic analysis aims to solve two problems: Small Parsimony Problem (SPP) and Big Parsimony Problem (BPP). Maximum parsimony is a popular approach for SPP and BPP which relies on itera- tively solving a NP-hard problem, the median problem. As a result, current median solvers and phylogenetic inference methods based on the median problem all face se- rious …

Cybersecurity Issues In The Context Of Cryptographic Shuffling Algorithms And Concept Drift: Challenges And Solutions, Hatim Alsuwat

Theses and Dissertations

In this dissertation, we investigate and address two kinds of data integrity threats. We first study the limitations of secure cryptographic shuffling algorithms regarding preservation of data dependencies. We then study the limitations of machine learning models regarding concept drift detection. We propose solutions to address these threats.

Shuffling Algorithms have been used to protect the confidentiality of sensitive data. However, these algorithms may not preserve data dependencies, such as functional de- pendencies and data-driven associations. We present two solutions for addressing these shortcomings: (1) Functional dependencies preserving shuffle, and (2) Data-driven asso- ciations preserving shuffle. For preserving functional dependencies, …

Challenges In Large-Scale Machine Learning Systems: Security And Correctness, Emad Alsuwat

Theses and Dissertations

In this research, we address the impact of data integrity on machine learning algorithms. We study how an adversary could corrupt Bayesian network structure learning algorithms by inserting contaminated data items. We investigate the resilience of two commonly used Bayesian network structure learning algorithms, namely the PC and LCD algorithms, against data poisoning attacks that aim to corrupt the learned Bayesian network model.

Data poisoning attacks are one of the most important emerging security threats against machine learning systems. These attacks aim to corrupt machine learning models by con- taminating datasets in the training phase. The lack of resilience of …

Semantic-Based Access Control Mechanisms In Dynamic Environments, Mouiad A. Hani Al-Wahah

Theses and Dissertations

The appearance of dynamic distributed networks in early eighties of the last century has evoked technologies like pervasive systems, ubiquitous computing, ambient intelligence, and more recently, Internet of Things (IoT) to be developed. Moreover, sensing capabil- ities embedded in computing devices offer users the ability to share, retrieve, and update resources on anytime and anywhere basis. These resources (or data) constitute what is widely known as contextual information. In these systems, there is an association between a system and its environment and the system should always adapt to its ever-changing environment. This situation makes the Context-Based Access Control (CBAC) the …

An Instruction Embedding Model For Binary Code Analysis, Kimberly Michelle Redmond

Theses and Dissertations

Binary code analysis is important for understanding programs without access to the original source code, which is common with proprietary software. Analyzing binaries can be challenging given their high variability: due to growth in tech manufactur- ers, source code is now frequently compiled for multiple instruction set architectures (ISAs); however, there is no formal dictionary that translates between their assem- bly languages. The difficulty of analysis is further compounded by different compiler optimizations and obfuscated malware signatures. Such minutiae means that some vulnerabilities may only be detectable on a fine-grained level. Recent strides in ma- chine learning—particularly in Natural Language …

Algorithms For Robot Coverage Under Movement And Sensing Constraints, Jeremy S. Lewis

Theses and Dissertations

This thesis explores the problem of generating coverage paths—that is, paths that pass within some sensor footprint of every point in an environment—for mobile robots. It both considers models for which navigation is a solved problem but motions are constrained, as well for models in which navigation must be considered along with coverage planning because of the robot’s unreliable sensing and movements.

The motion constraint we adopt for the former is a common constraint, that of a Dubins vehicle. We extend previous work that solves this coverage problem as a traveling salesman problem (TSP) by introducing a practical heuristic algorithm …

Interfacing Iconicity - Addressing Software Divarication Through Diagrammatic Design Principles, George Akhvlediani

Theses and Dissertations

This research examines conflicts accompanying the proliferation of computer technology and, more specifically, constellations of dependency in the always expanding volume of software, platforms, and the firms/individuals using them. We identify a pervasive phenomenon of “divarication” in the growing variety of progressively specialized systems and system roles. As software systems enter new thresholds of sophistication, they effectively aggregate many distinct components and protocols. Consequently, we are confronted with a diverse ecology of stratified and thereby incompatible software systems. Software inherits the limitations and potential flaws of its constituent parts, but unlike physical machinery, it isn’t readily disassembled in instances of …

Implementation Costs Of Spiking Versus Rate-Based Anns, Lacie Renee Stiffler

Theses and Dissertations

Artificial neural networks are an effective machine learning technique for a variety of data sets and domains, but exploiting the inherent parallelism in neural networks requires specialized hardware. Typically, computing the output of each neuron requires many multiplications, evaluation of a transcendental activation function, and transfer of its output to a large number of other neurons. These restrictions become more expensive when internal values are represented with increasingly higher data precision. A spiking neural network eliminates the limitations of typical rate-based neural networks by reducing neuron output and synapse weights to one-bit values, eliminating hardware multipliers, and simplifying the activation …

Uncertainty Estimation Of Deep Neural Networks, Chao Chen

Theses and Dissertations

Normal neural networks trained with gradient descent and back-propagation have received great success in various applications. On one hand, point estimation of the network weights is prone to over-fitting problems and lacks important uncertainty information associated with the estimation. On the other hand, exact Bayesian neural network methods are intractable and non-applicable for real-world applications. To date, approximate methods have been actively under development for Bayesian neural networks, including but not limited to: stochastic variational methods, Monte Carlo dropouts, and expectation propagation. Though these methods are applicable for current large networks, there are limits to these approaches with either underestimation …

Authenticating Users With 3d Passwords Captured By Motion Sensors, Jing Tian

Theses and Dissertations

Authentication plays a key role in securing various resources including corporate facilities or electronic assets. As the most used authentication scheme, knowledgebased authentication is easy to use but its security is bounded by how much a user can remember. Biometrics-based authentication requires no memorization but ‘resetting’ a biometric password may not always be possible. Thus, we propose study several behavioral biometrics (i.e., mid-air gestures) for authentication which does not have the same privacy or availability concerns as of physiological biometrics.

In this dissertation, we first propose a user-friendly authentication system Kin- Write that allows users to choose arbitrary, short and …