Physical Sciences and Mathematics Commons^™

An Efficient Privacy-Preserving Framework For Video Analytics, Tian Zhou

Doctoral Dissertations

With the proliferation of video content from surveillance cameras, social media, and live streaming services, the need for efficient video analytics has grown immensely. In recent years, machine learning based computer vision algorithms have shown great success in various video analytic tasks. Specifically, neural network models have dominated in visual tasks such as image and video classification, object recognition, object detection, and object tracking. However, compared with classic computer vision algorithms, machine learning based methods are usually much more compute-intensive. Powerful servers are required by many state-of-the-art machine learning models. With the development of cloud computing infrastructures, people are able …

Policy Gradient Methods: Analysis, Misconceptions, And Improvements, Christopher P. Nota

Doctoral Dissertations

Policy gradient methods are a class of reinforcement learning algorithms that optimize a parametric policy by maximizing an objective function that directly measures the performance of the policy. Despite being used in many high-profile applications of reinforcement learning, the conventional use of policy gradient methods in practice deviates from existing theory. This thesis presents a comprehensive mathematical analysis of policy gradient methods, uncovering misconceptions and suggesting novel solutions to improve their performance. We first demonstrate that the update rule used by most policy gradient methods does not correspond to the gradient of any objective function due to the way the …

Development Of A Decision Support System Webtool For Historic And Future Low Flow Estimation In The Northeast United States With Applications Of Machine Learning For Advancing Physical And Statistical Methodologies, Andrew F. Delsanto

Doctoral Dissertations

Droughts are a global challenge and anthropogenic climate change is expected to increase the frequency and severity of extreme low flow events. A major challenge for resource managers is how best to incorporate future climate change projections into low flow event estimations, especially in ungaged basins. Using both physically based hydrology models and statistical models, this dissertation contributes novel methodologies to three key challenges associated with 7-day, 10-year low flow (7Q10) estimation in the northeast United States. Chapter 2 builds upon statistically based 7Q10 estimation in ungaged basins by comparing multiple machine learning algorithms to classical statistical methodologies. This chapter’s …

Generalized Differentiable Neural Architecture Search With Performance And Stability Improvements, Emily J. Herron

Doctoral Dissertations

This work introduces improvements to the stability and generalizability of Cyclic DARTS (CDARTS). CDARTS is a Differentiable Architecture Search (DARTS)-based approach to neural architecture search (NAS) that uses a cyclic feedback mechanism to train search and evaluation networks concurrently, thereby optimizing the search process by enforcing that the networks produce similar outputs. However, the dissimilarity between the loss functions used by the evaluation networks during the search and retraining phases results in a search-phase evaluation network, a sub-optimal proxy for the final evaluation network utilized during retraining. ICDARTS, a revised algorithm that reformulates the search phase loss functions to ensure …

Generative Adversarial Game With Tailored Quantum Feature Maps For Enhanced Classification, Anais Sandra Nguemto Guiawa

Doctoral Dissertations

In the burgeoning field of quantum machine learning, the fusion of quantum computing and machine learning methodologies has sparked immense interest, particularly with the emergence of noisy intermediate-scale quantum (NISQ) devices. These devices hold the promise of achieving quantum advantage, but they grapple with limitations like constrained qubit counts, limited connectivity, operational noise, and a restricted set of operations. These challenges necessitate a strategic and deliberate approach to crafting effective quantum machine learning algorithms.

This dissertation revolves around an exploration of these challenges, presenting innovative strategies that tailor quantum algorithms and processes to seamlessly integrate with commercial quantum platforms. A …

Bayesian Structural Causal Inference With Probabilistic Programming, Sam A. Witty

Doctoral Dissertations

Reasoning about causal relationships is central to the human experience. This evokes a natural question in our pursuit of human-like artificial intelligence: how might we imbue intelligent systems with similar causal reasoning capabilities? Better yet, how might we imbue intelligent systems with the ability to learn cause and effect relationships from observation and experimentation? Unfortunately, reasoning about cause and effect requires more than just data: it also requires partial knowledge about data generating mechanisms. Given this need, our task then as computational scientists is to design data structures for representing partial causal knowledge, and algorithms for updating that knowledge in …

Enhancing The Performance Of The Mtcnn For The Classification Of Cancer Pathology Reports: From Data Annotation To Model Deployment, Kevin De Angeli

Doctoral Dissertations

Information contained in electronic health records (EHR) combined with the latest advances in machine learning (ML) have the potential to revolutionize the medical sciences. In particular, information contained in cancer pathology reports is essential to investigate cancer trends across the country. Unfortunately, large parts of information in EHRs are stored in the form of unstructured, free-text which limit their usability and research potential. To overcome this accessibility barrier, cancer registries depend on expert personnel who read, interpret, and extract relevant information. Naturally, as the number of stored pathology reports increases every day, depending on human experts presents scalability challenges. Recently, …

Transition Metal Computational Catalysis: Mechanistic Approaches And Development Of Novel Performance Metrics, Brett Anthony Smith

Doctoral Dissertations

Computational catalysis is an ever-growing field, thanks in part to the incredible progression of computational power and the efficiency offered by our current methodologies. Additionally, the accuracy of computation and the emergence of new methods that can decompose energetics and sterics into quantitative descriptors has allowed for researchers to begin to identify important structure-function relationships that predict the properties of unexplored subspaces within the overall chemical space. Catalytic descriptors have been used frequently in data driven high-throughput computational screenings. With the use of machine learning, a large portion of the chemical space an be predicted in matter of minutes or …

Better Understanding Genomic Architecture With The Use Of Applied Statistics And Explainable Artificial Intelligence, Jonathon C. Romero

Doctoral Dissertations

With the continuous improvements in biological data collection, new techniques are needed to better understand the complex relationships in genomic and other biological data sets. Explainable Artificial Intelligence (X-AI) techniques like Iterative Random Forest (iRF) excel at finding interactions within data, such as genomic epistasis. Here, the introduction of new methods to mine for these complex interactions is shown in a variety of scenarios. The application of iRF as a method for Genomic Wide Epistasis Studies shows that the method is robust in finding interacting sets of features in synthetic data, without requiring the exponentially increasing computation time of many …

Application Of Machine Learning In Geophysics: Ranking Teleseismic Shear Wave Splitting Measurements And Classifying Different Types Of Earthquakes, Yanwei Zhang

Doctoral Dissertations

"During the past decades, applications of Machine Learning have been explosively developed to solve various academic and industrial problems, and over-human performance has been shown in diverse areas. In geophysical research, Machine Learning, especially Convolutional Neural Network (CNN), has been applied in numerous studies and demonstrated considerable potential. In this study, we applied CNN to solve two geophysical problems, ranking teleseismic shear splitting (SWS) measurements and classifying different types of earthquakes.

For ranking teleseismic SWS measurements, we utilized a CNN-based method to automatically select reliable SWS measurements. The CNN was trained by human-verified teleseismic SWS measurements and tested using synthetic …

Pattern Formation And Phase Transition Of Connectivity In Two Dimensions, Arman Mohseni Kabir

Doctoral Dissertations

This dissertation is devoted to the study and analysis of different types of emergent behavior in physical systems. Emergence is a phenomenon that has fascinated researchers from various fields of science and engineering. From the emergence of global pandemics to the formation of reaction-diffusion patterns, the main feature that connects all these diverse systems is the appearance of a complex global structure as a result of collective interactions of simple underlying components. This dissertation will focus on two types of emergence in physical systems: emergence of long-range connectivity in networks and emergence and analysis of complex patterns. The most prominent …

Small Molecule Activation By Transition Metal Complexes: Studies With Quantum Mechanical And Machine Learning Methodologies, Justin Kyle Kirkland

Doctoral Dissertations

One of the largest areas of study in the fields of chemistry and engineering is that of activation of small molecules such as nitrogen, oxygen and methane. Herein we study the activation of such molecules by transition metal compounds using quantum mechanical methods in order to understand the complex chemistry behind these processes. By understanding these processes, we can design and propose novel catalytic species, and through the use of data-driven machine learning methods, we are able to accelerate materials discovery.

Machine Learning With Topological Data Analysis, Ephraim Robert Love

Doctoral Dissertations

Topological Data Analysis (TDA) is a relatively new focus in the fields of statistics and machine learning. Methods of exploiting the geometry of data, such as clustering, have proven theoretically and empirically invaluable. TDA provides a general framework within which to study topological invariants (shapes) of data, which are more robust to noise and can recover information on higher dimensional features than immediately apparent in the data. A common tool for conducting TDA is persistence homology, which measures the significance of these invariants. Persistence homology has prominent realizations in methods of data visualization, statistics and machine learning. Extending ML with …

Improving The Data Quality In Gravitation-Wave Detectors By Mitigating Transient Noise Artifacts, Kentaro Mogushi

Doctoral Dissertations

“The existence of gravitational waves (GWs), small perturbations in spacetime produced by accelerating massive objects was first predicted in 1916 as solutions of Einstein’s Theory of General Relativity (Einstein, 1916). Detecting and analyzing GWs produced by sources allows us to probe astrophysical phenomena.

The era of GW astronomy began from the first direct detection of the coalescence of a binary black hole in 2015 by the collaboration of the advanced Laser Interferometer Gravitational-wave Observatory (LIGO) (Aasi et al., 2015) and advanced Virgo (Abbott et al., 2016a). Since 2015, LIGO-Virgo detected about 50 confident transient events of GW signals (Abbott et …

Exposure Assessment Of Emerging Contaminants: Rapid Screening And Modeling Of Plant Uptake, Majid Bagheri

Doctoral Dissertations

"With the advent of new chemicals and their increasing uses in every aspect of our life, considerable number of emerging contaminants are introduced to environment yearly. Emerging contaminants in forms of pharmaceuticals, detergents, biosolids, and reclaimed wastewater can cross plant roots and translocate to various parts of the plants. Long-term human exposure to emerging contaminants through food consumption is assumed to be a pathway of interest. Thus, uptake and translocation of emerging contaminants in plants are important for the assessment of health risks associated with human exposure to emerging contaminants. To have a better understanding over fate of emerging contaminants …

Finding Critical And Gradient-Flat Points Of Deep Neural Network Loss Functions, Charles Gearhart Frye '09

Doctoral Dissertations

Despite the fact that the loss functions of deep neural networks are highly non-convex, gradient-based optimization algorithms converge to approximately the same performance from many random initial points. This makes neural networks easy to train, which, combined with their high representational capacity and implicit and explicit regularization strategies, leads to machine-learned algorithms of high quality with reasonable computational cost in a wide variety of domains.

One thread of work has focused on explaining this phenomenon by numerically characterizing the local curvature at critical points of the loss function, where gradients are zero. Such studies have reported that the loss functions …

Dynamic Composition Of Functions For Modular Learning, Clemens Gb Rosenbaum

Doctoral Dissertations

Compositionality is useful to reduce the complexity of machine learning models and increase their generalization capabilities, because new problems can be linked to the composition of existing solutions. Recent work has shown that compositional approaches can offer substantial benefits over a wide variety of tasks, from multi-task learning over visual question-answering to natural language inference, among others. A key variant is functional compositionality, where a meta-learner composes different (trainable) functions into complex machine learning models. In this thesis, I generalize existing approaches to functional compositionality under the umbrella of the routing paradigm, where trainable arbitrary functions are 'stacked' to form …

Studying The Effects Of Various Process Parameters On Early Age Hydration Of Single- And Multi-Phase Cementitious Systems, Rachel Cook

Doctoral Dissertations

”The hydration of multi-phase ordinary Portland cement (OPC) and its pure phase derivatives, such as tricalcium silicate (C₃S) and belite (ß-C₂S), are studied in the context varying process parameters -- for instance, variable water content, water activity, superplasticizer structure and dose, and mineral additive type and particle size. These parameters are studied by means of physical experiments and numerical/computational techniques, such as: thermodynamic estimations; numerical kinetic-based modelling; and artificial intelligence techniques like machine learning (ML) models. In the past decade, numerical kinetic modeling has greatly improved in terms of fitting experimental, isothermal calorimetry to kinetic-based modelling …

Machine Learning Methods For Personalized Health Monitoring Using Wearable Sensors, Annamalai Natarajan

Doctoral Dissertations

Mobile health is an emerging field that allows for real-time monitoring of individuals between routine clinical visits. Among others it makes it possible to remotely gather health signals, track disease progression and provide just-in-time interventions. Consumer grade wearable sensors can remotely gather health signals and other time series data. While wearable sensors can be readily deployed on individuals, there are significant challenges in converting raw sensor data into actionable insights. In this dissertation, we develop machine learning methods and models for personalized health monitoring using wearables. Specifically, we address three challenges that arise in these settings. First, data gathered from …

Machine Learning Techniques Implementation In Power Optimization, Data Processing, And Bio-Medical Applications, Khalid Khairullah Mezied Al-Jabery

Doctoral Dissertations

"The rapid progress and development in machine-learning algorithms becomes a key factor in determining the future of humanity. These algorithms and techniques were utilized to solve a wide spectrum of problems extended from data mining and knowledge discovery to unsupervised learning and optimization. This dissertation consists of two study areas. The first area investigates the use of reinforcement learning and adaptive critic design algorithms in the field of power grid control. The second area in this dissertation, consisting of three papers, focuses on developing and applying clustering algorithms on biomedical data. The first paper presents a novel modelling approach for …

Adaft: A Resource-Efficient Framework For Adaptive Fault-Tolerance In Cyber-Physical Systems, Ye Xu

Doctoral Dissertations

Cyber-physical systems frequently have to use massive redundancy to meet application requirements for high reliability. While such redundancy is required, it can be activated adaptively, based on the current state of the controlled plant. Most of the time the physical plant is in a state that allows for a lower level of fault-tolerance. Avoiding the continuous deployment of massive fault-tolerance will greatly reduce the workload of CPSs. In this dissertation, we demonstrate a software simulation framework (AdaFT) that can automatically generate the sub-spaces within which our adaptive fault-tolerance can be applied. We also show the theoretical benefits of AdaFT, and …

Data Analysis Methods Using Persistence Diagrams, Andrew Marchese

Doctoral Dissertations

In recent years, persistent homology techniques have been used to study data and dynamical systems. Using these techniques, information about the shape and geometry of the data and systems leads to important information regarding the periodicity, bistability, and chaos of the underlying systems. In this thesis, we study all aspects of the application of persistent homology to data analysis. In particular, we introduce a new distance on the space of persistence diagrams, and show that it is useful in detecting changes in geometry and topology, which is essential for the supervised learning problem. Moreover, we introduce a clustering framework directly …

Method For Enabling Causal Inference In Relational Domains, David Arbour

Doctoral Dissertations

The analysis of data from complex systems is quickly becoming a fundamental aspect of modern business, government, and science. The field of causal learning is concerned with developing a set of statistical methods that allow practitioners make inferences about unseen interventions. This field has seen significant advances in recent years. However, the vast majority of this work assumes that data instances are independent, whereas many systems are best described in terms of interconnected instances, i.e. relational systems. This discrepancy prevents causal inference techniques from being reliably applied in many real-world settings.
In this thesis, I will present three contributions to …

Intrinsic Functions For Securing Cmos Computation: Variability, Modeling And Noise Sensitivity, Xiaolin Xu

Doctoral Dissertations

A basic premise behind modern secure computation is the demand for lightweight cryptographic primitives, like identifier or key generator. From a circuit perspective, the development of cryptographic modules has also been driven by the aggressive scalability of complementary metal-oxide-semiconductor (CMOS) technology. While advancing into nano-meter regime, one significant characteristic of today's CMOS design is the random nature of process variability, which limits the nominal circuit design. With the continuous scaling of CMOS technology, instead of mitigating the physical variability, leveraging such properties becomes a promising way. One of the famous products adhering to this double-edged sword philosophy is the Physically …

Quantitative Metrics For Comparison Of Hyper-Dimensional Lsa Spaces For Semantic Differences, John Christopher Martin

Doctoral Dissertations

Latent Semantic Analysis (LSA) is a mathematically based machine learning technology that has demonstrated success in numerous applications in text analytics and natural language processing. The construction of a large hyper-dimensional space, a LSA space, is central to the functioning of this technique, serving to define the relationships between the information items being processed. This hyper-dimensional space serves as a semantic mapping system that represents learned meaning derived from the input content. The meaning represented in an LSA space, and therefore the mappings that are generated and the quality of the results obtained from using the space, is completely dependent …

Neuron Clustering For Mitigating Catastrophic Forgetting In Supervised And Reinforcement Learning, Benjamin Frederick Goodrich

Doctoral Dissertations

Neural networks have had many great successes in recent years, particularly with the advent of deep learning and many novel training techniques. One issue that has affected neural networks and prevented them from performing well in more realistic online environments is that of catastrophic forgetting. Catastrophic forgetting affects supervised learning systems when input samples are temporally correlated or are non-stationary. However, most real-world problems are non-stationary in nature, resulting in prolonged periods of time separating inputs drawn from different regions of the input space.

Reinforcement learning represents a worst-case scenario when it comes to precipitating catastrophic forgetting in neural networks. …

Automating Large-Scale Simulation Calibration To Real-World Sensor Data, Richard Everett Edwards

Doctoral Dissertations

Many key decisions and design policies are made using sophisticated computer simulations. However, these sophisticated computer simulations have several major problems. The two main issues are 1) gaps between the simulation model and the actual structure, and 2) limitations of the modeling engine's capabilities. This dissertation's goal is to address these simulation deficiencies by presenting a general automated process for tuning simulation inputs such that simulation output matches real world measured data. The automated process involves the following key components -- 1) Identify a model that accurately estimates the real world simulation calibration target from measured sensor data; 2) Identify …