Search results

Title: Face Recognition : Representation, Intrinsic Dimensionality, Capacity, and Demographic Bias
Creator: Gong, Sixue
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Face recognition is a widely adopted technology with numerous applications, such as mobile phone unlock, mobile payment, surveillance, social media and law enforcement. There has been tremendous progress in enhancing the accuracy of face recognition systems over the past few decades, much of which can be attributed to deep learning. Despite this progress, several fundamental problems in face recognition still remain unsolved. These problems include finding a salient representation, estimating...
Show moreFace recognition is a widely adopted technology with numerous applications, such as mobile phone unlock, mobile payment, surveillance, social media and law enforcement. There has been tremendous progress in enhancing the accuracy of face recognition systems over the past few decades, much of which can be attributed to deep learning. Despite this progress, several fundamental problems in face recognition still remain unsolved. These problems include finding a salient representation, estimating intrinsic dimensionality, representation capacity, and demographic bias. With growing applications of face recognition, the need for an accurate, robust, compact and fair representation is evident.In this thesis, we first develop algorithms to obtain practical estimates of intrinsic dimensionality of face representations, and propose a new dimensionality reduction method to project feature vectors from ambient space to intrinsic space. Based on the study in intrinsic dimensionality, we then estimate capacity of face representation, casting the face capacity estimation problem under the information theoretic framework of capacity of a Gaussian noise channel. Numerical experiments on unconstrained faces (IJB-C) provide a capacity upper bound of 27,000 for FaceNet and 84,000 for SphereFace representation at 1% FAR. In the second part of the thesis, we address the demographic bias problem in face recognition systems where errors are lower on certain cohorts belonging to specific demographic groups. We propose two de-biasing frameworks that extract feature representations to improve fairness in face recognition. Experiments on benchmark face datasets (RFW, LFW, IJB-A, and IJB-C) show that our approaches are able to mitigate face recognition bias on various demographic groups (biasness drops from 6.83 to 5.07) as well as maintain the competitive performance (i.e., 99.75% on LFW, and 93.70% TAR @ 0.1% FAR on IJB-C). Lastly, we explore the global distribution of deep face representations derived from correlations between image samples of within-class and cross-class to enhance the discriminativeness of face representation of each identity in the embedding space. Our new approach to face representation achieves state-of-the-art performance for both verification and identification tasks on benchmark datasets (99.78% on LFW, 93.40% on CPLFW, 98.41% on CFP-FP, 96.2% TAR @ 0.01% FAR and 95.3% Rank-1 accuracy on IJB-C). Since, the primary techniques we employ in this dissertation are not specific to faces only, we believe our research can be extended to other problems in computer vision, for example, general image classification and representation learning.
Show less

Title: Evolving Phenotypically Plastic Digital Organisms
Creator: Lalejini, Alexander
Date: 2021
Collection: Electronic Theses & Dissertations
Description: The ability to dynamically respond to cues from the environment is a fundamental feature of most adaptive systems. In biological systems, changes to an organism based on environmental cues is called phenotypic plasticity. Indeed, phenotypic plasticity underlies many of the adaptive traits and developmental patterns found in nature and serves as a key mechanism for responding to spatially or temporally variable environments. Most computer programs require phenotypic plasticity, as they must...
Show moreThe ability to dynamically respond to cues from the environment is a fundamental feature of most adaptive systems. In biological systems, changes to an organism based on environmental cues is called phenotypic plasticity. Indeed, phenotypic plasticity underlies many of the adaptive traits and developmental patterns found in nature and serves as a key mechanism for responding to spatially or temporally variable environments. Most computer programs require phenotypic plasticity, as they must respond dynamically to stimuli such as user input, sensor data, et cetera. As such, phenotypic plasticity also has practical applications in genetic programming, wherein we apply the natural principles of evolution to automatically synthesize computer programs rather than writing them by hand. In this dissertation, I achieve two synergistic aims: (1) I use populations of self-replicating computer programs (digital organisms) to empirically study the conditions under which adaptive phenotypic plasticity evolves and how its evolution shapes subsequent evolutionary outcomes; and (2) I transfer insights from biology to develop novel genetic programming techniques in order to evolve more responsive (i.e., phenotypically plastic) computer programs. First, I illustrate the importance of mutation rate, environmental change, and partially-plastic building blocks for the evolution of adaptive plasticity. Next, I show that adaptive phenotypic plasticity stabilizes populations against environmental change, allowing them to more easily retain novel adaptive traits. Finally, I improve our ability to evolve phenotypically plastic computer programs with three novel genetic programming techniques: (1) SignalGP, which provides mechanisms to control code expression based on environmental cues, (2) tag-based genetic regulation to adjust code expression based on current context, and (3) tag-accessed memory to provide more dynamic mechanisms for storing data.
Show less

Title: Online Learning Algorithms for Mining Trajectory data and their Applications
Creator: Wang, Ding
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Trajectories are spatio-temporal data that represent traces of moving objects, such as humans, migrating animals, vehicles, and tropical cyclones. In addition to the geo-location information, a trajectory data often contain other (non-spatial) features describing the states of the moving objects. The time-varying geo-location and state information would collectively characterize a trajectory dataset, which can be harnessed to understand the dynamics of the moving objects. This thesis focuses...
Show moreTrajectories are spatio-temporal data that represent traces of moving objects, such as humans, migrating animals, vehicles, and tropical cyclones. In addition to the geo-location information, a trajectory data often contain other (non-spatial) features describing the states of the moving objects. The time-varying geo-location and state information would collectively characterize a trajectory dataset, which can be harnessed to understand the dynamics of the moving objects. This thesis focuses on the development of efficient and accurate machine learning algorithms for forecasting the future trajectory path and state of a moving object. Although many methods have been developed in recent years, there are still numerous challenges that have not been sufficiently addressed by existing methods, which hamper their effectiveness when applied to critical applications such as hurricane prediction. These challenges include their difficulties in terms of handling concept drifts, error propagation in long-term forecasts, missing values, and nonlinearities in the data. In this thesis, I present a family of online learning algorithms to address these challenges. Online learning is an effective approach as it can efficiently fit new observations while adapting to concept drifts present in the data. First, I proposed an online learning framework called OMuLeT for long-term forecasting of the trajectory paths of moving objects. OMuLeT employs an online learning with restart strategy to incrementally update the weights of its predictive model as new observation data become available. It can also handle missing values in the data using a novel weight renormalization strategy.Second, I introduced the OOR framework to predict the future state of the moving object. Since the state can be represented by ordinal values, OOR employs a novel ordinal loss function to train its model. In addition, the framework was extended to OOQR to accommodate a quantile loss function to improve its prediction accuracy for larger values on the ordinal scale. Furthermore, I also developed the OOR-ε and OOQR-ε frameworks to generate real-valued state predictions using the ε insensitivity loss function.Third, I developed an online learning framework called JOHAN, that simultaneously predicts the location and state of the moving object. JOHAN generates its predictions by leveraging the relationship between the state and location information. JOHAN utilizes a quantile loss function to bias the algorithm towards predicting more accurately large categorical values in terms of the state of the moving object, say, for a high intensity hurricane.Finally, I present a deep learning framework to capture non-linear relationships in trajectory data. The proposed DTP framework employs a TDM approach for imputing missing values, coupled with an LSTM architecture for dynamic path prediction. In addition, the framework was extended to ODTP, which applied an online learning setting to address concept drifts present in the trajectory data.As proof of concept, the proposed algorithms were applied to the hurricane prediction task. Both OMuLeT and ODTP were used to predict the future trajectory path of a hurricane up to 48 hours lead time. Experimental results showed that OMuLeT and ODTP outperformed various baseline methods, including the official forecasts produced by the U.S. National Hurricane Center. OOR was applied to predict the intensity of a hurricane up to 48 hours in advance. Experimental results showed that OOR outperformed various state-of-the-art online learning methods and can generate predictions close to the NHC official forecasts. Since hurricane intensity prediction is a notoriously hard problem, JOHAN was applied to improve its prediction accuracy by leveraging the trajectory information, particularly for high intensity hurricanes that are near landfall.
Show less

Title: 5D Nondestructive Evaluation : Object Reconstruction to Toolpath Generation
Creator: Hamilton, Ciaron Nathan
Date: 2021
Collection: Electronic Theses & Dissertations
Description: The focus of this thesis is to provide virtualization methods for a Cyber-Physical System (CPS) setup that interfaces physical Nondestructive Evaluation (NDE) scanning environments into virtual spaces through virtual-physical interfacing and path planning. In these environments, a probe used for NDE mounted as the end-effector of a robot arm will actuate and acquire data along the surface of a Material Under Test (MUT) within virtual and physical spaces. Such configurations are practical for...
Show moreThe focus of this thesis is to provide virtualization methods for a Cyber-Physical System (CPS) setup that interfaces physical Nondestructive Evaluation (NDE) scanning environments into virtual spaces through virtual-physical interfacing and path planning. In these environments, a probe used for NDE mounted as the end-effector of a robot arm will actuate and acquire data along the surface of a Material Under Test (MUT) within virtual and physical spaces. Such configurations are practical for applications that require damage analysis of certain geometrically complex parts, ranging from automobile to aerospace to military industries. The pipeline of the designed $5D$ actuation system starts by virtually reconstructing the physical MUT and its surrounding environment, generating a toolpath along the surface of the reconstructed MUT, conducting a physical scan along the toolpath which synchronizes the robot's end effector position with retrieved NDE data, and post processing the obtained data. Most of this thesis will focus on virtual topics, including reconstruction from stereo camera images and toolpath planning. Virtual mesh generation of the MUT and surrounding environment are found with stereo camera images, where methods for camera positioning, registration, filtering, and reconstruction are provided. Path planning around the MUT uses a customized path-planner, where a $2D$ grid of rays is generated where each ray intersection across the surface of the MUT's mesh provides the translation and rotation of waypoints for actuation. Experimental setups include both predefined meshes and reconstructed meshes found from several real carbon-fiber automobile components using an Intel RealSense D425i stereo camera, showing both the reconstruction and path planning results. A theoretical review is also included to discuss analytical prospects of the system. The final system is designed to be automated to minimize human interaction to conduct scans, with later reports planned to discuss the scanning and post processing prospects of the system.
Show less

Title: COMBINING FACE AND IRIS FOR PRIVACY PRESERVATION
Creator: Ledala, Achsah Junia
Date: 2021
Collection: Electronic Theses & Dissertations
Description: With the extensive use of biometrics for authenticating users, the need to ensure privacy of biometric data is greater than ever before. Biometric authentication systems are vulnerable to attacks and the loss of biometric data will lead to loss of privacy of an individual. Multibiometrics refers to the use of multiple biometric modalities simultaneously in order to perform matching. In this work, we introduce a multibiometric fusion technique which can be used to ensure that the original raw...
Show moreWith the extensive use of biometrics for authenticating users, the need to ensure privacy of biometric data is greater than ever before. Biometric authentication systems are vulnerable to attacks and the loss of biometric data will lead to loss of privacy of an individual. Multibiometrics refers to the use of multiple biometric modalities simultaneously in order to perform matching. In this work, we introduce a multibiometric fusion technique which can be used to ensure that the original raw biometric data are unlikely to be compromised and, at the same time, recognition can be performed. The face and the iris biometric modalities are fused at the feature-level to produce discriminative embeddings that can be used for recognition. The original face or the iris cannot be retrieved from the combined representation, thus preserving the privacy of the individual. We present the results of this approach, provide analysis, discuss the challenges, and list possible future directions.
Show less

Title: OPTIMIZATION OF LARGE SCALE ITERATIVE EIGENSOLVERS
Creator: Afibuzzaman, Md
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Sparse matrix computations, in the form of solvers for systems of linear equations, eigenvalue problem or matrix factorizations constitute the main kernel in problems from fields as diverse as computational fluid dynamics, quantum many body problems, machine learning and graph analytics. Iterative eigensolvers have been preferred over the regular method because the regular method not being feasible with industrial sized matrices. Although dense linear algebra libraries like BLAS, LAPACK,...
Show moreSparse matrix computations, in the form of solvers for systems of linear equations, eigenvalue problem or matrix factorizations constitute the main kernel in problems from fields as diverse as computational fluid dynamics, quantum many body problems, machine learning and graph analytics. Iterative eigensolvers have been preferred over the regular method because the regular method not being feasible with industrial sized matrices. Although dense linear algebra libraries like BLAS, LAPACK, SCALAPACK are well established and some vendor optimized implementation like mkl from Intel or Cray Libsci exist, it is not the same case for sparse linear algebra which is lagging far behind. The main reason behind slow progress in the standardization of sparse linear algebra or library development is the different forms and properties depending on the application area. It is worsened for deep memory hierarchies of modern architectures due to low arithmetic intensities and memory bound computations. Minimization of data movement and fast access to the matrix are critical in this case. Since the current technology is driven by deep memory architectures where we get the increased capacity at the expense of increased latency and decreased bandwidth when we go further from the processors. The key to achieve high performance in sparse matrix computations in deep memory hierarchy is to minimize data movement across layers of the memory and overlap data movement with computations. My thesis work contributes towards addressing the algorithmic challenges and developing a computational infrastructure to achieve high performance in scientific applications for both shared memory and distributed memory architectures. For this purpose, I started working on optimizing a blocked eigensolver and optimized specific computational kernels which uses a new storage format. Using this optimization as a building block, we introduce a shared memory task parallel framework focusing on optimizing the entire solvers rather than a specific kernel. Before extending this shared memory implementation to a distributed memory architecture, I simulated the communication pattern and overheads of a large scale distributed memory application and then I introduce the communication tasks in the framework to overlap communication and computation. Additionally, I also tried to find a custom scheduler for the tasks using a graph partitioner. To get acquainted with high performance computing and parallel libraries, I started my PhD journey with optimizing a DFT code named Sky3D where I used dense matrix libraries. Despite there might not be any single solution for this problem, I tried to find an optimized solution. Though the large distributed memory application MFDn is kind of the driver project of the thesis, but the framework we developed is not confined to MFDn only, rather it can be used for other scientific applications too. The output of this thesis is the task parallel HPC infrastructure that we envisioned for both shared and distributed memory architectures.
Show less

Title: Contributions to Fingerprint Recognition
Creator: Engelsma, Joshua James
Date: 2021
Collection: Electronic Theses & Dissertations
Description: From the early days of the mid to late nineteenth century when scientific research first began to focus on fingerprints, to the present day fingerprint recognition systems we find deployed on our day to day devices, the science of fingerprint recognition has come a long way. In spite of this progress, there remains challenging problems to be solved. This thesis highlights a few of these problems, and proposes solutions to address them. One area of further research that must be conducted on...
Show moreFrom the early days of the mid to late nineteenth century when scientific research first began to focus on fingerprints, to the present day fingerprint recognition systems we find deployed on our day to day devices, the science of fingerprint recognition has come a long way. In spite of this progress, there remains challenging problems to be solved. This thesis highlights a few of these problems, and proposes solutions to address them. One area of further research that must be conducted on fingerprint recognition systems is that of robust, operational evaluations. In chapter two of this thesis, we show how the current practices of using calibration patterns to evaluate fingerprint readers are limited. We then propose a realistic fake finger called the Universal Target. The Universal Target is a realistic, 3D, fake finger (or phantom) which can be imaged by all major types of fingerprint sensing technologies. We show the entire manufacturing (molding and casting) process for fabricating the Universal Targets. Then, we show a series of evaluations which demonstrate how the Universal Targets can be used to operationally evaluate current commercial fingerprint readers. Our Universal Target is a significant step forward in enabling more realistic, standardized evaluations of fingerprint readers. In our third chapter, we shift gears from improving the evaluation standards of fingerprint readers to instead focus on the security of fingerprint readers. In particular, we turn our attention towards detecting fake fingerprint (spoof) attacks. To do so, we open source a fingerprint reader (built from low-cost ubiquitous components), called RaspiReader. RaspiReader is a high-resolution fingerprint reader customized with both direct-view imaging and FTIR imaging in order to better detect fingerprint spoofs. We show through a number of experiments that RaspiReader enables state-of-the-art fingerprint spoof detection accuracy. We also demonstrate that RaspiReader enables better generalization to what are known as "unseen attacks" (those attacks which were not seen during training of the spoof detector). Finally, we show that fingerprints captured by RaspiReader are completely compatible with images captured by legacy fingerprint readers for matching.In chapter four, we move on to propose a major improvement to the fingerprint feature extraction and matching sub-modules of fingerprint recognition systems. In particular, we propose a deep network, called DeepPrint, to extract a 200 byte fixed-length fingerprint representation. While prevailing fingerprint matchers primarily utilize minutiae points and expensive graph matching algorithms for comparison, two DeepPrint representations can be compared with only 192 multiplications and 191 additions. This is extremely useful for large scale search where potentially billions of pairwise fingerprint comparisons must be made. The DeepPrint representation also enables practical encrypted matching using a fully homomorphic encryption scheme. This enables better protection of the fingerprint templates which are stored in the database. While discriminative fixed-length representations are available for both face and iris recognition, such a representation has eluded fingerprint recognition. This chapter aims to fill that void.Finally, we conclude our thesis by working to extend fingerprint recognition to all ages. While current fingerprint recognition systems are being used by billions of teenagers and adults around the world, the youngest people among us remain disenfranchised. In particular, modern day fingerprint recognition systems do not work well on infants and young children. In this penultimate chapter, we aim to rectify this major shortcoming. To that end, we prototype a high-resolution (1900 ppi) infant fingerprint reader. Then, we track and fingerprint 315 infants (under the age of 3 months at enrollment) at the Dayalbagh Children's Hospital in Agra India over the course of 1 year (4 different sessions). To match the infant fingerprints, we develop our own high-resolution infant fingerprint matcher. Our experimental results demonstrate significant promise for the extension of fingerprint recognition to all ages. This work has the potential for major global good as all young infants and children could be given a verifiable digital identity for better vaccination tracking as a child and for government benefits and assistance as an adult. In summary, this thesis makes major contributions to the entire end-to-end fingerprint recognition system and extends its use case to all ages.
Show less

Title: SIGN LANGUAGE RECOGNIZER FRAMEWORK BASED ON DEEP LEARNING ALGORITHMS
Creator: Akandeh, Atra
Date: 2021
Collection: Electronic Theses & Dissertations
Description: According to the World Health Organization (WHO, 2017), 5% of the world’s population have hearing loss. Most people with hearing disabilities communicate via sign language, which hearing people find extremely difficult to understand. To facilitate communication of deaf and hard of hearing people, developing an efficient communication system is a necessity. There are many challenges associated with the Sign Language Recognition (SLR) task, namely, lighting conditions, complex background,...
Show moreAccording to the World Health Organization (WHO, 2017), 5% of the world’s population have hearing loss. Most people with hearing disabilities communicate via sign language, which hearing people find extremely difficult to understand. To facilitate communication of deaf and hard of hearing people, developing an efficient communication system is a necessity. There are many challenges associated with the Sign Language Recognition (SLR) task, namely, lighting conditions, complex background, signee body postures, camera position, occlusion, complexity and large variations in hand posture, no word alignment, coarticulation, etc.Sign Language Recognition has been an active domain of research since the early 90s. However, due to computational resources and sensing technology constraints, limited advancement has been achieved over the years. Existing sign language translation systems mostly can translate a single sign at a time, which makes them less effective in daily-life interaction. This work develops a novel sign language recognition framework using deep neural networks, which directly maps videos of sign language sentences to sequences of gloss labels by emphasizing critical characteristics of the signs and injecting domain-specific expert knowledge into the system. The proposed model also allows for combining data from variant sources and hence combating limited data resources in the SLR field.
Show less

Title: Coreference Resolution for Downstream NLP Tasks
Creator: Pani, Sushanta Kumar
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Natural Language Processing (NLP) tasks have witnessed a significant improvement in performance by utilizing the power of end-to-end neural network models. An NLP system built for one job can contribute to other closely related tasks. Coreference Resolution (CR) systems work on resolving references and are at the core of many NLP tasks. The coreference resolution refers to the linking of repeated object references in a text. CR systems can boost the performance of downstream NLP tasks, such...
Show moreNatural Language Processing (NLP) tasks have witnessed a significant improvement in performance by utilizing the power of end-to-end neural network models. An NLP system built for one job can contribute to other closely related tasks. Coreference Resolution (CR) systems work on resolving references and are at the core of many NLP tasks. The coreference resolution refers to the linking of repeated object references in a text. CR systems can boost the performance of downstream NLP tasks, such as Text Summarization, Question Answering, Machine Translation, etc. We provide a detailed comparative error analysis of two state-of-the-art coreference resolution systems to understand error distribution in the predicted output. The understanding of error distribution is helpful to interpret the system behavior. Eventually, this will contribute to the selection of an optimal CR system for a specific target task.
Show less

Title: Towards Robust and Secure Face Recognition : Defense Against Physical and Digital Attacks
Creator: Deb, Debayan
Date: 2021
Collection: Electronic Theses & Dissertations
Description: The accuracy, usability, and touchless acquisition of state-of-the-art automated face recognition systems (AFR) have led to their ubiquitous adoption in a plethora of domains, including mobile phone unlock, access control systems, and payment services. Despite impressive recognition performance, prevailing AFR systems remain vulnerable to the growing threat of face attacks which can be launched in both physical and digital domains. Face attacks can be broadly classified into three attack...
Show moreThe accuracy, usability, and touchless acquisition of state-of-the-art automated face recognition systems (AFR) have led to their ubiquitous adoption in a plethora of domains, including mobile phone unlock, access control systems, and payment services. Despite impressive recognition performance, prevailing AFR systems remain vulnerable to the growing threat of face attacks which can be launched in both physical and digital domains. Face attacks can be broadly classified into three attack categories: (i) Spoof attacks: artifacts in the physical domain (e.g., 3D masks, eye glasses, replaying videos), (ii) Adversarial attacks: imperceptible noises added to probes for evading AFR systems, and (iii) Digital manipulation attacks: entirely or partially modified photo-realistic faces using generative models. Each of these categories is composed of different attack types. For example, each spoof medium, e.g., 3D mask and makeup, constitutes one attack type. Likewise, in adversarial and digital manipulation attacks, each attack model, designed by unique objectives and losses, may be considered as one attack type. Thus, the attack categories and types form a 2-layer tree structure encompassing the diverse attacks. Such a tree will inevitably grow in the future. Given the growing dissemination of ``fake news” and "deepfakes", the research community and social media platforms alike are pushing towards generalizable defense against continuously evolving and sophisticated face attacks. In this dissertation, we first propose a set of defense methods that achieve state-of-the-art performance in detecting attack types within individual attack categories, both physical (e.g., face spoofs) and digital (e.g., adversarial faces and digital manipulation), then introduce a method for simultaneously safeguarding against each attack.First, in an effort to impart generalizability and interpretability to face spoof detection systems, we propose a new face anti-spoofing framework specifically designed to detect unknown spoof types, namely, Self-Supervised Regional Fully Convolutional Network (SSR-FCN), that is trained to learn local discriminative cues from a face image in a self-supervised manner. The proposed framework improves generalizability while maintaining the computational efficiency of holistic face anti-spoofing approaches (< 4 ms on a Nvidia GTX 1080Ti GPU). The proposed method is also interpretable since it localizes which parts of the face are labeled as spoofs. Experimental results show that SSR-FCN can achieve True Detection Rate (TDR) = 65% @ 2.0% False Detection Rate (FDR) when evaluated on a dataset comprising of 13 different spoof types under unknown attacks while achieving competitive performances under standard benchmark face anti-spoofing datasets (Oulu-NPU, CASIA-MFSD, and Replay-Attack).Next, we address the problem of defending against adversarial attacks. We first propose, AdvFaces, an automated adversarial face synthesis method that learns to generate minimal perturbations in the salient facial regions. Once AdvFaces is trained, it can automatically evade state-of-the-art face matchers with attack success rates as high as 97.22% and 24.30% at 0.1% FAR for obfuscation and impersonation attacks, respectively. We then propose a new self-supervised adversarial defense framework, namely FaceGuard, that can automatically detect, localize, and purify a wide variety of adversarial faces without utilizing pre-computed adversarial training samples. FaceGuard automatically synthesizes diverse adversarial faces, enabling a classifier to learn to distinguish them from bona fide faces. Concurrently, a purifier attempts to remove the adversarial perturbations in the image space. FaceGuard can achieve 99.81%, 98.73%, and 99.35% detection accuracies on LFW, CelebA, and FFHQ, respectively, on six unseen adversarial attack types.Finally, we take the first steps towards safeguarding AFR systems against face attacks in both physical and digital domains. We propose a new unified face attack detection framework, namely UniFAD, which automatically clusters similar attacks and employs a multi-task learning framework to learn salient features to distinguish between bona fides and coherent attack types. The proposed UniFAD can detect face attacks from 25 attack types across all 3 attack categories with TDR = 94.73% @ 0.2% FDR on a large fake face dataset, namely GrandFake. Further, UniFAD can identify whether attacks are adversarial, digitally manipulated, or contain spoof artifacts, with 97.37% classification accuracy.
Show less

Title: Optimal Learning of Deployment and Search Strategies for Robotic Teams
Creator: Wei, Lai
Date: 2021
Collection: Electronic Theses & Dissertations
Description: In the problem of optimal learning, the dilemma of exploration and exploitation stems from the fact that gathering information and exploiting it are, in many cases, two mutually exclusive activities. The key to optimal learning is to strike a balance between exploration and exploitation. The Multi-Armed Bandit (MAB) problem is a prototypical example of such an explore-exploit tradeoff, in which a decision-maker sequentially allocates a single resource by repeatedly choosing one among a set of...
Show moreIn the problem of optimal learning, the dilemma of exploration and exploitation stems from the fact that gathering information and exploiting it are, in many cases, two mutually exclusive activities. The key to optimal learning is to strike a balance between exploration and exploitation. The Multi-Armed Bandit (MAB) problem is a prototypical example of such an explore-exploit tradeoff, in which a decision-maker sequentially allocates a single resource by repeatedly choosing one among a set of options that provide stochastic rewards. The MAB setup has been applied in many robotics problems such as foraging, surveillance, and target search, wherein the task of robots can be modeled as collecting stochastic rewards. The theoretical work of this dissertation is based on the MAB setup and three problem variations, namely heavy-tailed bandits, nonstationary bandits, and multi-player bandits, are studied. The first two variations capture two key features of stochastic feedback in complex and uncertain environments: heavy-tailed distributions and nonstationarity; while the last one addresses the problem of achieving coordination in uncertain environments. We design several algorithms that are robust to heavy-tailed distributions and nonstationary environments. Besides, two distributed policies that require no communication among agents are designed for the multi-player stochastic bandits in a piece-wise stationary environment.The MAB problems provide a natural framework to study robotic search problems. The above variations of the MAB problems directly map to robotic search tasks in which a robot team searches for a target from a fixed set of view-points (arms). We further focus on the class of search problems involving the search of an unknown number of targets in a large or continuous space. We view the multi-target search problem as a hot-spots identification problem in which, instead of the global maximum of the field, all locations with a value greater than a threshold need to be identified. We consider a robot moving in 3D space with a downward-facing camera sensor. We model the robot's sensing output using a multi-fidelity Gaussian Process (GP) that systematically describes the sensing information available at different altitudes from the floor. Based on the sensing model, we design a novel algorithm that (i) addresses the coverage-accuracy tradeoff: sampling at a location farther from the floor provides a wider field of view but less accurate measurements, (ii) computes an occupancy map of the floor within a prescribed accuracy and quickly eliminates unoccupied regions from the search space, and (iii) travels efficiently to collect the required samples for target detection. We rigorously analyze the algorithm and establish formal guarantees on the target detection accuracy and the detection time.An approach to extend the single robot search policy to multiple robots is to partition the environment into multiple regions such that workload is equitably distributed among all regions and then assign a robot to each region. The coverage control focuses on such equitable partitioning and the workload is equivalent to the so-called service demands in the coverage control literature. In particular, we study the adaptive coverage control problem, in which the demands of robotic service within the environment are modeled as a GP. To optimize the coverage of service demands in the environment, the team of robots aims to partition the environment and achieve a configuration that minimizes the coverage cost, which is a measure of the average distance of a service demand from the nearest robot. The robots need to address the explore-exploit tradeoff: to minimize coverage cost, they need to gather information about demands within the environment, whereas information gathering deviates them from maintaining a good coverage configuration. We propose an algorithm that schedules learning and coverage epochs such that its emphasis gradually shifts from exploration to exploitation while never fully ceasing to learn. Using a novel definition of coverage regret, we analyze the algorithm and characterizes its coverage performance over a finite time horizon.
Show less

Title: The Evolution of Fundamental Neural Circuits for Cognition in Silico
Creator: Tehrani-Saleh, Ali
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Despite decades of research on intelligence and fundamental components of cognition, we still know very little about the structure and functionality of nervous systems. Questions in cognition and intelligent behavior are addressed by scientists in the fields of behavioral biology, neuroscience, psychology, and computer science. Yet it is difficult to reverse engineer observed sophisticated intelligent behaviors in animals and even more difficult to understand their underlying mechanisms.In...
Show moreDespite decades of research on intelligence and fundamental components of cognition, we still know very little about the structure and functionality of nervous systems. Questions in cognition and intelligent behavior are addressed by scientists in the fields of behavioral biology, neuroscience, psychology, and computer science. Yet it is difficult to reverse engineer observed sophisticated intelligent behaviors in animals and even more difficult to understand their underlying mechanisms.In this dissertation, I use a recently-developed neuroevolution platform -called Markov brain networks- in which Darwinian selection is used to evolve both structure and functionality of digital brains. I use this platform to study some of the most fundamental cognitive neural circuits: 1) visual motion detection, 2) collision-avoidance based on visual motion cues, 3) sound localization, and 4) time perception. In particular, I investigate both the selective pressures and environmental conditions in the evolution of these cognitive components, as well as the circuitry and computations behind them. This dissertation lays the groundwork for an evolutionary agent-based method to study the neural circuits for cognition in silico.
Show less

Title: Replaying Life's Virtual Tape : Examining the Role of History in Experiments with Digital Organisms
Creator: Bundy, Jason Nyerere
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Evolution is a complex process with a simple recipe. Evolutionary change involves three essential “ingredients” interacting over many generations: adaptation (selection), chance (random variation), and history (inheritance). In 1989’s Wonderful Life, the late paleontologist Stephen Jay Gould advocated for the importance of historical contingency—the way unique events throughout history influence future possibilities—using a clever thought experiment of “replaying life’s tape”. But not...
Show moreEvolution is a complex process with a simple recipe. Evolutionary change involves three essential “ingredients” interacting over many generations: adaptation (selection), chance (random variation), and history (inheritance). In 1989’s Wonderful Life, the late paleontologist Stephen Jay Gould advocated for the importance of historical contingency—the way unique events throughout history influence future possibilities—using a clever thought experiment of “replaying life’s tape”. But not everyone was convinced. Some believed that chance was the primary driver of evolutionary change, while others insisted that natural selection was the most powerful influence. Since then, “replaying life’s tape” has become a core method in experimental evolution for measuring the relative contributions of adaptation, chance, and history. In this dissertation, I focus on the effects associated with history in evolving populations of digital organisms—computer programs that self-replicate, mutate, compete, and evolve in virtual environments. In Chapter 1, I discuss the philosophical significance of Gould’s thought experiment and its influence on experimental methods. I argue that his thought experiment was a challenge to anthropocentric reasoning about natural history that is still popular, particularly outside of the scientific community. In this regard, it was his way of advocating for a “radical” view of evolution. In Chapter 2—Richard Lenski, Charles Ofria, and I describe a two-phase, virtual, “long-term” evolution experiment with digital organisms using the Avida software. In Phase I, we evolved 10 replicate populations, in parallel, from a single genotype for around 65,000 generations. This part of the experiment is similar to the design of Lenski’s E. coli Long-term Evolution Experiment (LTEE). We isolated the dominant genotype from each population around 3,000 generations (shallow history) into Phase I and then again at the end of Phase I (deep history). In Phase II, we evolved 10 populations from each of the genotypes we isolated from Phase I in two new environments, one similar and one dissimilar to the old environment used for Phase I. Following Phase II, we estimated the contributions of adaptation, chance, and history to the evolution of fitness and genome length in each new environment. This unique experimental design allowed us to see how the contributions of adaptation, chance, and history changed as we extended the depth of history from Phase I. We were also able to determine whether the results depended on the extent of environmental change (similar or dissimilar new environment). In Chapter 3, we report an extended analysis of the experiment from the previous chapter to further examine how extensive adaptation to the Phase I environment shaped the evolution of replicates during Phase II. We show how the form of pleiotropy (antagonistic or synergistic) between the old (Phase I) and new (Phase II) habitats was influenced by the depth of history from Phase I (shallow or deep) and the extent of environmental change (similar or dissimilar new environment). In the final chapter Zachary Blount, Richard Lenski, and I describe an exercise we developed using the educational version of Avida (Avida-ED). The exercise features a two-phase, “replaying life’s tape” activity. Students are able to explore how the unique history of founders that we pre-evolved during Phase I influences the acquisition of new functions by descendent populations during Phase II, which the students perform during the activity.
Show less

Title: CS1 AND GENDER : UNDERSTANDING EFFECTS OF BACKGROUND AND SELF- EFFICACY ON ACHIEVEMENT AND INTEREST
Creator: Sands, Philip
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Over the past 20 years, the field of computer science has experienced a growth in studentinterest. Despite this increase in participation rates, longstanding gender gaps persist in computer science. Recent research has examined a wide variety of individual factors (e.g., self-efficacy, sense of belonging, etc.) that impact student interest and achievement in computer science; however, these factors are rarely considered in the context of existing learning theories. In this correlational study...
Show moreOver the past 20 years, the field of computer science has experienced a growth in studentinterest. Despite this increase in participation rates, longstanding gender gaps persist in computer science. Recent research has examined a wide variety of individual factors (e.g., self-efficacy, sense of belonging, etc.) that impact student interest and achievement in computer science; however, these factors are rarely considered in the context of existing learning theories. In this correlational study, I explored the relationship between prior knowledge of computer programming, self-efficacy, and the sources of self-efficacy as they differed by gender in a theoretical model of achievement and interest for students in first-year computer science (CS1) courses. This model was based on prior work from Bandura (1997) and others exploring self- efficacy and social cognitive theory in the context of mathematics and science fields. Using cross-sectional data from N=182 CS1 students at two universities, structural regressions were conducted between factors impacting CS1 students across the entire population and for men (N=108) and women (N=70) individually. This data was then used to address the following research questions. (1A) How do prior knowledge of computer programming, the sources of self- efficacy, and self-efficacy for computing predict CS1 achievement and student intentions to continue study in CS? (1B) How does self-efficacy mediate the relationship between student prior knowledge of computer programming and achievement in CS1? (1C) How are thoserelationships moderated by gender? (2) How does feedback in the form of student grades impact intention to continue in CS when considering gender as a moderating factor? For all students, student self-efficacy for CS positively impacted CS1 achievement and post-CS1 interest. Aligning with past research, self-efficacy was derived largely from mastery experiences, with vicarious experiences and social persuasions also contributing to a moderate degree. Social persuasions had a negative effect on self-efficacy, which diverged from research in other fields. The relationship between prior knowledge of computer programming and CS1 achievement was not mediated by self-efficacy and had a small positive effect. For women, vicarious experiences played a stronger role in defining student self-efficacy in CS. Additionally, while the importance of self-efficacy on achievement was similar to that for men, self-efficacy and achievement both played a much stronger role in determining student interest in CS for women. All these findings are in need of further exploration as the analysis was underpowered due to a small, COVID-19 impacted sample size. Future work should focus on the role of feedback on student self-efficacy, the potential misalignment of CS1 feedback and social network feedback, and interventions that address student beliefs about CS abilities to increase opportunities for authentic mastery and vicarious experiences.
Show less

Title: Automated Speaker Recognition in Non-ideal Audio Signals Using Deep Neural Networks
Creator: Chowdhury, Anurag
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Speaker recognition entails the use of the human voice as a biometric modality for recognizing individuals. While speaker recognition systems are gaining popularity in consumer applications, most of these systems are negatively affected by non-ideal audio conditions, such as audio degradations, multi-lingual speech, and varying duration audio. This thesis focuses on developing speaker recognition systems robust to non-ideal audio conditions.Firstly, a 1-Dimensional Convolutional Neural...
Show moreSpeaker recognition entails the use of the human voice as a biometric modality for recognizing individuals. While speaker recognition systems are gaining popularity in consumer applications, most of these systems are negatively affected by non-ideal audio conditions, such as audio degradations, multi-lingual speech, and varying duration audio. This thesis focuses on developing speaker recognition systems robust to non-ideal audio conditions.Firstly, a 1-Dimensional Convolutional Neural Network (1D-CNN) is developed to extract noise-robust speaker-dependent speech characteristics from the Mel Frequency Cepstral Coefficients (MFCC). Secondly, the 1D-CNN-based approach is extended to develop a triplet-learning-based feature-fusion framework, called 1D-Triplet-CNN, for improving speaker recognition performance by judiciously combining MFCC and Linear Predictive Coding (LPC) features. Our hypothesis rests on the observation that MFCC and LPC capture two distinct aspects of speech: speech perception and speech production. Thirdly, a time-domain filterbank called DeepVOX is learned from vast amounts of raw speech audio to replace commonly-used hand-crafted filterbanks, such as the Mel filterbank, in speech feature extractors. Finally, a vocal style encoding network called DeepTalk is developed to learn speaker-dependent behavioral voice characteristics to improve speaker recognition performance. The primary contribution of the thesis is the development of deep learning-based techniques to extract discriminative, noise-robust physical and behavioral voice characteristics from non-ideal speech audio. A large number of experiments conducted on the TIMIT, NTIMIT, SITW, NIST SRE (2008, 2010, and 2018), Fisher, VOXCeleb, and JukeBox datasets convey the efficacy of the proposed techniques and their importance in improving speaker recognition performance in non-ideal audio conditions.
Show less

Title: TEACHERS IN SOCIAL MEDIA : A DATA SCIENCE PERSPECTIVE
Creator: Karimi, Hamid
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Social media has become an integral part of human life in the 21st century. The number of social media users was estimated to be around 3.6 billion individuals in 2020. Social media platforms (e.g., Facebook) have facilitated interpersonal communication, diffusion of information, the creation of groups and communities, to name a few. As far as education systems are concerned, online social media has transformed and connected traditional social networks within the schoolhouse to a broader and...
Show moreSocial media has become an integral part of human life in the 21st century. The number of social media users was estimated to be around 3.6 billion individuals in 2020. Social media platforms (e.g., Facebook) have facilitated interpersonal communication, diffusion of information, the creation of groups and communities, to name a few. As far as education systems are concerned, online social media has transformed and connected traditional social networks within the schoolhouse to a broader and expanded world outside. In such an expanded virtual space, teachers engage in various activities within their communities, e.g., exchanging instructional resources, seeking new teaching methods, engaging in online discussions. Therefore, given the importance of teachers in social media and its tremendous impact on PK-12 education, in this dissertation, we investigate teachers in social media from a data science perspective. Our investigation in this direction is essentially an interdisciplinary endeavor bridging modern data science and education. In particular, we have made three contributions, as briefly discussed in the following. Current teachers in social media studies suffice to a small number of surveyed teachers while thousands of other teachers are on social media. This hinders us from conducting large-scale data-driven studies pertinent to teachers in social media. Aiming to overcome this challenge and further facilitate data-driven studies related to teachers in social media, we propose a novel method that automatically identifies teachers on Pinterest, an image-based social media popular among teachers. In this framework, we formulate the teacher identification problem as a positive unlabelled (PU) learning where positive samples are surveyed teachers, and unlabelled samples are their online friends. Using our framework, we build the largest dataset of teachers on Pinterest. With this dataset at our disposal, we perform an exploratory analysis of teachers on Pinterest while considering their genders. Our analysis incorporates two crucial aspects of teachers in social media. First, we investigate various online activities of male and female teachers, e.g., topics and sources of their curated resources, the professional language employed to describe their resources. Second, we investigate male and female teachers in the context of the social network (the graph) they belong to, e.g., structural centrality, gender homophily. Our analysis and findings in this part of the dissertation can serve as a valuable reference for many entities concerned with teachers' gender, e.g., principals, state, and federal governments.Finally, in the third part of the dissertation, we shed light on the diffusion of teacher-curated resources on Pinterest. First, we introduce three measures to characterize the diffusion process. Then, we investigate these three measures while considering two crucial characteristics of a resource, e.g., the topic and the source. Ultimately, we investigate how teacher attributes (e.g., the number of friends) affect the diffusion of their resources. The conducted diffusion analysis is the first of its kind and offers a deeper understating of the complex mechanism driving the diffusion of resources curated by teachers on Pinterest.
Show less

Title: Computational Frameworks for Indel-Aware Evolutionary Analysis using Large-Scale Genomic Sequence Data
Creator: Wang, Wei
Date: 2021
Collection: Electronic Theses & Dissertations
Description: With the development of sequencing techniques, genetic sequencing data has been extensively used in evolutionary studies.The phylogenetic reconstruction problem, which is the reconstruction of evolutionary history from biomolecular sequences, is a fundamental problem. The evolutionary relationship between organisms is often represented by phylogeny, which is a tree or network representation. The most widely-used approach for reconstructing phylogenies from sequencing data involves two phases:...
Show moreWith the development of sequencing techniques, genetic sequencing data has been extensively used in evolutionary studies.The phylogenetic reconstruction problem, which is the reconstruction of evolutionary history from biomolecular sequences, is a fundamental problem. The evolutionary relationship between organisms is often represented by phylogeny, which is a tree or network representation. The most widely-used approach for reconstructing phylogenies from sequencing data involves two phases: multiple sequence alignment and phylogenetic reconstruction from the aligned sequences. As the amount of biomolecular sequence data increases, it has become a major challenge to develop efficient and accurate computational methods for phylogenetic analyses of large-scale sequencing data. Due to the complexity of the phylogenetic reconstruction problem in modern phylogenetic studies, the traditional sequence-based phylogenetic analysis methods involve many over-simplified assumptions. In this thesis, we describe our contribution in relaxing some of these over-simplified assumptions in the phylogenetic analysis.Insertion and deletion events, referred to as indels, carry much phylogenetic information but are often ignored in the reconstruction process of phylogenies. We take into account the indel uncertainties in multiple phylogenetic analyses by applying resampling and re-estimation. Another over-simplified assumption that we contributed to is adopted by many commonly used non-parametric algorithms for the resampling of biomolecular sequences, all sites in an MSA are evolved independently and identically distributed (i.i.d). Many evolution events, such as recombination and hybridization, may produce intra-sequence and functional dependence in biomolecular sequences that violate this assumption. We introduce SERES, a resampling algorithm for biomolecular sequences that can produce resampled replicates that preserve the intra-sequence dependence. We describe the application of the SERES resampling and re-estimation approach to two classical problems: the multiple sequence alignment support estimation and recombination-aware local genealogical inference. We show that these two statistical inference problems greatly benefit from the indel-aware resampling and re-estimation approach and the reservation of intra-sequence dependence.A major drawback of SERES is that it requires parameters to ensure the synchronization of random walks on unaligned sequences.We introduce RAWR, a non-parametric resampling method designed for phylogenetic tree support estimation that does not require extra parameters. We show that the RAWR-based resampling and re-estimation method produces comparable or typically better performance than the traditional bootstrap approach on the phylogenetic tree support estimation problem. We further relax the commonly used assumption of phylogeny.Evolutionary history is usually considered as a tree structure. Evolutionary events that cause reticulated gene flow are ignored. Previous studies show that alignment uncertainty greatly impacts downstream tree inference and learning. However, there is little discussion about the impact of MSA uncertainties on the phylogenetic network reconstruction. We show evidence that the errors introduced in MSA estimation decrease the accuracy of the inferred phylogenetic network, and an indel-aware reconstruction method is needed for phylogenetic network analysis. In this dissertation, we introduce our contribution to phylogenetic estimation using biomolecular sequence data involving complex evolutionary histories, such as sequence insertion and deletion processes and non-tree-like evolution.
Show less

Title: EXTENDED REALITY (XR) & GAMIFICATION IN THE CONTEXT OF THE INTERNET OF THINGS (IOT) AND ARTIFICIAL INTELLIGENCE (AI)
Creator: Pappas, Georgios
Date: 2021
Collection: Electronic Theses & Dissertations
Description: The present research develops a holistic framework for and way of thinking about Deep Technologies related to Gamification, eXtended Reality (XR), the Internet of Things (IoT), and Artificial Intelligence (AI). Starting with the concept of gamification and the immersive technology of XR, we create interconnections with the IoT and AI implementations. While each constituent technology has its own unique impact, our approach uniquely addresses the combinational potential of these technologies...
Show moreThe present research develops a holistic framework for and way of thinking about Deep Technologies related to Gamification, eXtended Reality (XR), the Internet of Things (IoT), and Artificial Intelligence (AI). Starting with the concept of gamification and the immersive technology of XR, we create interconnections with the IoT and AI implementations. While each constituent technology has its own unique impact, our approach uniquely addresses the combinational potential of these technologies that may have greater impact than any technology on its own. To approach the research problem more efficiently, the methodology followed includes its initial division into smaller parts. For each part of the research problem, novel applications were designed and developed including gamified tools, serious games and AR/VR implementations. We apply the proposed framework in two different domains: autonomous vehicles (AVs), and distance learning.Specifically, in chapter 2, an innovative hybrid tool for distance learning is showcased where, among others, the fusion with IoT provides a novel pseudomultiplayer mode. This mode may transform advanced asynchronous gamified tools to synchronous by enabling or disabling virtual events and phenomena enhancing the student experience. Next, in Chapter 3, along with gamification, the combination of XR with IoT data streams is presented but this time in an automotive context. We showcase how this fusion of technologies provides low-latency monitoring of vehicle characteristics, and how this can be visualized in augmented and virtual reality using low-cost hardware and services. This part of our proposed framework provides the methodology of creating any type of Digital Twin with near real-time data visualization.Following that, in chapter 4 we establish the second part of the suggested holistic framework where Virtual Environments (VEs), in general, can work as synthetic data generators and thus, be a great source of artificial suitable for training AI models. This part of the research includes two novel implementations the Gamified Digital Simulator (GDS) and the Virtual LiDAR Simulator.Having established the holistic framework, in Chapter 5, we now “zoom in” to gamification exploring deeper aspects of virtual environments and discuss how serious games can be combined with other facets of virtual layers (cyber ranges,virtual learning environments) to provide enhanced training and advanced learning experiences. Lastly, in chapter 6, “zooming out” from gamification an additional enhancement layer is presented. We showcase the importance of human-centered design of via an implementation that tries to simulate the AV-pedestrian interactions in a virtual and safe environment.
Show less

Title: The Evolutionary Origins of Cognition : Understanding the early evolution of biological control systems and general intelligence
Creator: Carvalho Pontes, Anselmo
Date: 2021
Collection: Electronic Theses & Dissertations
Description: In the last century, we have made great strides towards understanding natural cognition and recreating it artificially. However, most cognitive research is still guided by an inadequate theoretical framework that equates cognition to a computer system executing a data processing task. Cognition, whether natural or artificial, is not a data processing system; it is a control system.At cognition's core is a value system that allows it to evaluate current conditions and decide among two or more...
Show moreIn the last century, we have made great strides towards understanding natural cognition and recreating it artificially. However, most cognitive research is still guided by an inadequate theoretical framework that equates cognition to a computer system executing a data processing task. Cognition, whether natural or artificial, is not a data processing system; it is a control system.At cognition's core is a value system that allows it to evaluate current conditions and decide among two or more courses of action. Memory, learning, planning, and deliberation, rather than being essential cognitive abilities, are features that evolved over time to support the primary task of deciding “what to do next”. I used digital evolution to recreate the early stages in the evolution of natural cognition, including the ability to learn. Interestingly, I found cognition evolves in a predictable manner, with more complex abilities evolving in stages, by building upon previous simpler ones. I initially investigated the evolution of dynamic foraging behaviors among the first animals known to have a central nervous system, Ediacaran microbial mat miners. I then followed this up by evolving more complex forms of learning. I soon encountered practical limitations of the current methods, including exponential demand of computational resources and genetic representations that were not conducive to further scaling. This type of complexity barrier has been a recurrent issue in digital evolution. Nature, however, is not limited in the same ways; through evolution, it has created a language to express robust, modular, and flexible control systems of arbitrary complexity and apparently open-ended evolvability. The essential features of this language can be captured in a digital evolution platform. As an early demonstration of this, I evolved biologically plausible regulatory systems for virtual cyanobacteria. These systems regulate the cells' growth, photosynthesis and replication given the daily light cycle, the cell's energy reserves, and levels of stress. Although simple, this experimental system displays dynamics and decision-making mechanisms akin to biology, with promising potential for open-ended evolution of cognition towards general intelligence.
Show less

Title: Deep Convolutional Networks for Modeling Geo-Spatio-Temporal Relationships and Extremes
Creator: Wilson, Tyler
Date: 2021
Collection: Electronic Theses & Dissertations
Description: Geo-spatio-temporal data are valuable for a broad range of applications including traffic forecasting, weather prediction, detection of epidemic outbreaks, and crime monitoring. Data driven approaches to these problems must address several fundamental challenges such as handling the %The two we focus on are the importance ofgeo-spatio-temporal relationships and extreme events. Another recent technological shift has been the success of deep learning especially in applications such as computer...
Show moreGeo-spatio-temporal data are valuable for a broad range of applications including traffic forecasting, weather prediction, detection of epidemic outbreaks, and crime monitoring. Data driven approaches to these problems must address several fundamental challenges such as handling the %The two we focus on are the importance ofgeo-spatio-temporal relationships and extreme events. Another recent technological shift has been the success of deep learning especially in applications such as computer vision, speech recognition, and natural language processing. In this work, we argue that deep learning is a promising approach for many geo-spatio-temporal problems and highlight how it can be used to address the challenges of modeling geo-spatio-temporal relationships and extremes. Though previous research has established techniques for modeling spatio-temporal relationships, these approaches are often limited to gridded spatial data with fixed-length feature vectors and considered only spatial relationships among the features, while ignoring the relationships among model parameters.We begin by describing how the spatial and temporal relationships for non-gridded spatial data can be modeled simultaneously by coupling the graph convolutional network with a long short-term memory (LSTM) network. Unlike previous research, our framework treats the adjacency matrix associated with the spatial data as a model parameter that can be learned from data, with constraints on its sparsity and rank to reduce the number of estimated parameters.Further, we show that the learned adjacency matrix may reveal useful information about the dominant spatial relationships that exist within the data. Second, we explore the varieties of spatial relationships that may exist in a geo-spatial prediction task. Specifically, we distinguish between spatial relationships among predictors and the spatial relationships among model parameters at different locations. We demonstrate an approach for modeling spatial dependencies among model parameters using graph convolution and provide guidance on when convolution of each type can be effectively applied. We evaluate our proposed approach on a climate downscaling and weather prediction tasks. Next, we introduce DeepGPD, a novel deep learning framework for predicting the distribution of geo-spatio-temporal extreme events. We draw on research in extreme value theory and use the generalized Pareto distribution (GPD) to model the distribution of excesses over a threshold. The GPD is integrated into our deep learning framework to learn the distribution of future excess values while incorporating the geo-spatio-temporal relationships present in the data. This requires a novel reparameterization of the GPD to ensure that its constraints are satisfied by the outputs of the neural network. We demonstrate the effectiveness of our proposed approach on a real-world precipitation data set. DeepGPD also employs a deep set architecture to handle the variable-sized feature sets corresponding to excess values from previous time steps as its predictors. Finally, we extend the DeepGPD formulation to simultaneously predict the distribution of extreme events and accurately infer their point estimates. Doing so requires modeling the full distribution of the data not just its extreme values. We propose DEMM, a deep mixture model for modeling the distribution of both excess and non-excess values. To ensure the point estimation of DEMM is a feasible value, new constraints on the output of the neural network are introduced, which requires a new reparameterization of the model parameters of the GPD. We conclude by discussing possibilities for further research at the intersection of deep learning and geo-spatio-temporal data.
Show less

MSU Libraries

Digital Repository

Enabled Filters

Pages

Pages

MSU Libraries

Digital Repository

You are here

Search results

Pages

Pages