Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
105 results
Search Results
Item SENSORY AND HORMONAL MECHANISMS OF EARLY LIFE BEHAVIOR IN A SOCIAL CICHLID FISH(2024) Westbrook, Molly; Juntti, Scott; Biology; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Studying the ontogeny of animal behavior is fundamental to ethology and allows understanding how behaviors in early life may affect later life success. The social cichlid Astatotilapia burtoni is an excellent model for examining the mechanisms of early life aggression due to the robust social hierarchy enforced by stereotyped, measurable social behaviors. We examine how hormonal signaling affects early life aggression through pharmacology and CRISPR-Cas9 mutants. We test which sensory pathways convey aggression-eliciting stimuli through sensory deprivation experiments. And we identify kinematic features that predict aggression through machine-learning video tracking algorithms. We observe that aggressive behaviors emerge around 17 days post fertilization (dpf), correlating with when the animals transition to free swimming away from the mother. We find that sex steroids subtly organize behavioral circuits for aggression and suggest that unknown additional mechanisms play a leading role. We show that thyroid hormone is not necessary or sufficient for the transition to aggressive behavior. We show that visual signals are necessary for the full expression of aggression, but in the absence of visual signal, low levels of aggression remain. We show that ciliated olfactory receptor signaling maintains low levels of aggression, as mutant animals display higher levels of aggressive behavior between 17 and 24 dpf. Finally, we demonstrate that swimming velocity has potential to predict aggressive instances of behavior. Together, we find multiple levels of control for early life aggressive bouts from sensory input to hormonal organization of brain circuits.Item Analyzing Inverse Design Problems from a Topological Perspective(2024) Chen, Qiuyi; Fuge, Mark; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Inverse design (ID) problems are inverse problems that aim to rapidly retrieve the subset of valid designs having the desired performances and properties under the given conditions. In practice, this can be solved by training generative models to approximate and sample the posterior distributions of designs. However, little has been done to understand their mechanisms and limitations from a theoretical perspective. This dissertation leverages theoretical tools from general and differential topology to answer these three questions of inverse design: what does a set of valid designs look like? How helpful are the data-driven generative models for retrieving the desired designs from this set? What topological properties affect the subset of desired designs? The dissertation proceeds by dismantling inverse (design) problems into two major subjects: that is, the representing and probing of a given set of valid designs (or data), and the retrieval of the desired designs (or data) from this given set. It draws inspiration from topology and geometry to investigate them and makes the main contributions below: 1. Chapter 3 details a novel representation learning method called Least Volume, which has properties similar to nonlinear PCA for representing datasets. It can minimize the representation's dimension automatically and, as shown in Chapter 4, conducts contrastive learning when applied to labeled datasets. 2. Two conditional generative models are developed to generate performant 2-D airfoils and 3-D heat sinks in Chapter 5 and 6 respectively. They can produce realistic designs to warm-start further optimization, with the relevant chapters detailing their acceleration effects. 3. Lastly, Chapter 7 describes how to use Least volume to solve high-dimensional inverse problems efficiently. Specifically, using examples from physic system identification, the chapter uncovers the correlation between the inverse problem's uncertainty and its intrinsic dimensions.Item Deployment of Large Vision and Language Models for Real-Time Robotic Triage in a Mass Casualty Incident(2024) Mangel, Alexandra Paige; Paley, Derek; Aerospace Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In the event of a mass casualty incident, such as a natural disaster or war zone, having a system of triage in place that is efficient and accurate is critical for life-saving intervention, but medical personnel and resources are often strained and struggle to provide immediate care to those in need. This thesis proposes a system of autonomous air and ground vehicles equipped with stand-off sensing equipment designed to detect and localize casualties and assess them for critical injury patterns. The goal is to assist emergency medical technicians in identifying those in need of primary care by using generative AI models to analyze casualty images and communicate with the victims. Large language models are explored for the purpose of developing a chatbot that can ask a casualty where they are experiencing pain and make an informed assessment about injury classifications, and a vision language model is prompt engineered to assess a casualty image to produce a report on designated injury classifiers.Item Object-Attribute Compositionality for Visual Understanding(2024) Saini, Nirat; Shrivastava, Abhinav Dr; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Object appearances evolve overtime, which results in visually discernible changes in their colors, shapes, sizes and materials. Humans are innately good at recognizing and understanding the evolution of object states, which is also crucial for visual understanding across images and videos. However, current vision models still struggle to capture and account for these subtle changes to recognize the objects and underlying action causing the changes. This thesis focuses on using compositional learning for recognition and generation of attribute-object pairs. In the first part, we propose to disentangle visual features for object and attributes, to generalize recognition for novel object-attribute pairs. Next, we extend this approach to learn entirely unseen attribute-object pairs, by using semantic language priors, label smoothing and propagation techniques. Further, we use object states for action recognition in videos where subtle changes in object attributes and affordances help in identifying state-modifying and context-transforming actions. All of these methods for decomposing and composing objects and states generalize to unseen pairs and out-of-domain datasets for various compositional zero-shot learning and action recognition tasks. In the second part, we propose a new benchmark suite Chop \& Learn for a novel task of Compositional Image Generation as well as discuss the implications of these approaches for other compositional tasks in images, videos, and beyond. We further extend insertion and editing of attributes of objects consistently across frames of videos, using off-the-shelf training free architecture and discuss the future challenges and opportunities of compositionality for visual understanding.Item Machine Learning with Differentiable Physics Priors(2024) Qiao, Yiling; Lin, Ming ML; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Differentiable physics priors enable gradient-based learning systems to adhere to physical dynamics. By making physics simulations differentiable, we can backpropagate through the physical consequences of actions. This pipeline allows agents to quickly learn to achieve desired effects in the physical world and is an effective technique for solving inverse problems in physical or dynamical systems. This new programming paradigm bridges model-based and data-driven methods, mitigating data scarcity and model bias simultaneously. My research focuses on developing scalable, powerful, and efficient differentiable physics simulators. We have created state-of-the-art differentiable physics for rigid bodies, cloth, fluids, articulated bodies, and deformable solids, achieving performance orders of magnitude better than existing alternatives. These differentiable simulators are applied to solve inverse problems, train control policies, and enhance reinforcement learning algorithms.Item Efficient Optimization Algorithms for Nonconvex Machine Learning Problems(2024) Xian, Wenhan; Huang, Heng HH; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In recent years, the success of the AI revolution has led to the training of larger neural networks on vast amounts of data to achieve superior performance. These powerful machine learning models have enabled the creation of remarkable AI products. Optimization, as the core of machine learning, becomes especially crucial because most machine learning problems can ultimately be formulated as optimization problems, which require minimizing a loss function with respect to model parameters based on training samples. To enhance the efficiency of optimization algorithms, distributed learning has emerged as a popular solution for addressing large-scale machine learning tasks. In distributed learning, multiple worker nodes collaborate to train a global model. However, a key challenge in distributed learning is the communication cost. This thesis introduces a novel adaptive gradient algorithm with gradient sparsification to address this issue. Another significant challenge in distributed learning is the communication overhead on the central parameter server. To mitigate this bottleneck, decentralized distributed (serverless) learning has been proposed, where each worker node only needs to communicate with its neighbors. This thesis investigates core nonconvex optimization problems in decentralized settings, including constrained optimization, minimax optimization, and second-order optimality. Efficient optimization algorithms are proposed to solve these problems. Additionally, the convergence analysis of minimax optimization under the generalized smooth condition is explored. A generalized algorithm is proposed, which can be applied to a broader range of applications.Item Analyzing the Dynamics of Biological and Artificial Neural Networks with Applications to Machine Learning(2024) Srinivasan, Keshav; Girvan, Michelle; Biophysics (BIPH); Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)The study of the brain has profoundly shaped the evolution of computational learning models and the history of neural networks. This journey began in the 1940s with Warren McCulloch and Walter Pitts’ groundbreaking work on the first mathematical model of a neuron, laying the foundation for artificial neural networks. The 1950s and 60s witnessed a significant milestone with Frank Rosenblatt’s development of the perceptron, showcasing the potential of neural networks for complex computational tasks. Since then, the field of neural networks has witnessed explosive growth, and terms like “Artificial Intelligence” and “Machine Learning” have become commonplace across diverse fields, including finance,medicine, and science. This dissertation explores the symbiotic parallels between neuroscience and machine learning, focusing on the dynamics of biological and artificial neural networks. We begin by examining artificial neural networks, particularly in predicting the dynamics of large, complex networks—a paradigm where traditional machine learning algorithms often struggle. To address this, we propose a novel approach utilizing a parallel architecture that mimics the network’s structure, achieving scalable and accurate predictions. Shifting our focus to biological neuronal networks, we delve into the theory of critical systems. This theory posits that the brain, when viewed as a complex dynamical system, operates near a critical point, a state ideal for efficient information processing. A key experimental observation of this type of criticality is neuronal avalanches—scale-free cascades of neuronal activity—which have been documented both in vitro (in neuronal cultures and acute brain slices) and in vivo (in the brains of awake animals). Recent advancements in experimental techniques, such as multi-photon imaging and genetically encoded fluorescent markers, allow for the measurement of activity in living organisms with unparalleled single-cell resolution. Despite these advances, significant challenges remain when only a fraction of neurons can be recorded with sufficient resolution, leading to inaccurate estimations of power-law relationships in size, duration, and scaling of neuronal avalanches. We demonstrate that by analyzing simulated critical neuronal networks alongside real 2-photon imaging data, temporal coarse-graining can recover the critical value of the mean size vs. duration scaling of neuronal avalanches, allowing for more accurate estimations of critical brain dynamics even from subsampled data. Finally, we bridge the gap between machine learning and neuroscience by exploring the concept of excitatory-inhibitory balance, a crucial feature of neuronal networks in the brain, within the framework of reservoir computing. We emphasize the stabilizing role of inhibition in reservoir computers (RCs), mirroring its function in the brain. We propose a novel inhibitory adaptation mechanism that allows RCs to autonomously adjust inhibitory connections to achieve a specific firing rate target, motivated by the firing rate homeostasis observed in biological neurons. Overall, this dissertation strives to deepen the ongoing collaboration between neuroscience and machine learning, fostering advancements that will benefit both fields.Item Understanding and Enhancing Machine Learning Models with Theoretical Foundations(2024) Hu, Zhengmian; Huang, Heng HH; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Machine learning has become a key driver of many contemporary technological advancements. With its empirical success, there is an urgent need for theoretical research to explain and complement these practical achievements. This includes understanding the empirical success of machine learning, especially deep learning, and aiding the design of better algorithms in terms of performance, efficiency, and security. This dissertation aims to advance the understanding and practical development of machine learning through three interrelated research directions, while emphasizing reliable theoretical guarantees throughout the process. In the first part, we study the deep learning theory under overparameterization conditions. The core objects of study are the Conjugate Kernel and Neural Tangent Kernel, which have deep connections to the training dynamics of deep learning. Based on the analysis of these kernels, we prove several new concentration results characterizing the trainability and generalization of infinitely wide neural networks. In the second part, we focus on training algorithms. On one hand, we propose new algorithms to improve learning efficiency. This includes a new underdamped Langevin MCMC method called ALUM, for which we prove its complexity reaches the theoretical lower bound. On the other hand, we propose new theoretical tools to analyze existing algorithms and obtain tighter convergence results. For Proxskip, our analysis shows it can still achieve an improvement in communication complexity from sublinear to linear convergence under stochastic oracle. We also generalize the concept of Lipschitz smoothness for tighter non-convex optimization analysis. In the third part, we develop new Monte Carlo methods to large language models (LLMs) to improve their efficiency and security. We develop unbiased watermarking techniques to protect model outputs and propose an Accelerated Speculative Sampling method for faster inference. We also investigate the trade-off between watermark strength and inference sampling efficiency, pointing out the conflict between the two.Item Studies in Differential Privacy and Federated Learning(2024) Zawacki, Christopher Cameron; Abed, Eyad H; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In the late 20th century, Machine Learning underwent a paradigm shift from model-driven to data-driven design. Rather than field specific models, advances in sensors, data storage, and computing power enabled the collection of increasing amounts of data. The abundance of new data allowed researchers to fit flexible models directly to observed data. The influx of information made possible numerous advances, including the development of novel medicines, increases in efficiency of markets, and the proliferation of vast sensor networks. However, not all data should be freely accessible. Sensitive medical records, personal finances, and private IDs are all currently stored on digital devices across the world with the expectation that they remain private. However, at the same time, such data is frequently instrumental in the development of predictive models. Since the beginning of the 21st century, researchers have recognized that traditional methods of anonymizing data are inadequate for protecting client identities. This dissertation's primary focus is the advancement of two fields of data privacy: Differential Privacy and Federated Learning. Differential Privacy is one of the most successful modern privacy methods. By injecting carefully structured noise into a dataset, Differential Privacy obscures individual contributions while allowing researchers to extract meaningful information from the aggregate. Within this methodology, the Gaussian mechanism is one of the most common privacy mechanisms due to its favorable properties such as the ability of each client to apply noise locally before transmission to a server. However, the use of this mechanism yields only an approximate form of Differential Privacy. This dissertation introduces the first in-depth analysis of the Symmetric alpha-Stable (SaS) privacy mechanism, demonstrating its ability to achieve pure-Differential Privacy while retaining local applicability. Based on these findings, the dissertation advocates for using the SaS privacy mechanism in protecting the privacy of client data. Federated Learning is a sub-field of Machine Learning, which trains Machine Learning models across a collection (federation) of client devices. This approach aims to protect client privacy by limiting the type of information that clients transmit to the server. However, this distributed environment poses challenges such as non-uniform data distributions and inconsistent client update rates, which reduces the accuracy of trained models. To overcome these challenges, we introduce Federated Inference, a novel algorithm that we show is consistent in federated environments. That is, even when the data is unevenly distributed and the clients' responses to the server are staggered in time (asynchronous), the algorithm is able to converge to the global optimum. We also present a novel result in system identification in which we extend a method known as Dynamic Mode Decomposition to accommodate input delayed systems. This advancement enhances the accuracy of identifying and controlling systems relevant to privacy-sensitive applications such as smart grids and autonomous vehicles. Privacy is increasingly pertinent, especially as investments in computer infrastructure constantly grow in order to cater to larger client bases. Privacy failures impact an ever-growing number of individuals. This dissertation reports on our efforts to advance the toolkit of data privacy tools through novel methods and analysis while navigating the challenges of the field.Item OPTIMAL PROBING OF BATTERY CYCLES FOR MACHINE LEARNING-BASED MODEL DEVELOPMENT(2024) Nozarijouybari, Zahra; Fathy, Hosam HF; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This dissertation examines the problems of optimizing the selection of the datasets and experiments used for parameterizing machine learning-based electrochemical battery models. The key idea is that data selection, or “probing” can empower such models to achieve greater fidelity levels. The dissertation is motivated by the potential of battery models to enable theprediction and optimization of battery performance and control strategies. The literature presents multiple battery modeling approaches, including equivalent circuit, physics-based, and machine learning models. Machine learning is particularly attractive in the battery systems domain, thanks to its flexibility and ability to model battery performance and aging dynamics. Moreover, there is a growing interest in the literature in hybrid models that combine the benefits of machine learning with either the simplicity of equivalent circuit models or the predictiveness of physics-based models or both. The focus of this dissertation is on both hybrid and purely data-driven battery models. Moreover, the overarching question guiding the dissertation is: how does the selection of the datasets and experiments used for parameterizing these models affect their fidelity and parameter identifiability? Parameter identifiability is a fundamental concept from information theory that refers to the degree to which one can accurately estimate a given model’s parameters from input-output data. There is substantial existing research in the literature on battery parameter identifiability. An important lesson from this literature is that the design of a battery experiment can affect parameter identifiability significantly. Hence, test trajectory optimization has the potential to substantially improve model parameter identifiability. The literature explores this lesson for equivalent circuit and physics-based battery models. However, there is a noticeable gap in the literature regarding identifiability analysis and optimization for both machine learning-based and hybrid battery models. To address the above gap, the dissertation makes four novel contributions to the literature. The first contribution is an extensive survey of the machine learning-based battery modeling literature, highlighting the critical need for information-rich and clean datasets for parameterizing data-driven battery models. The second contribution is a K-means clustering-based algorithm for detecting outlier patterns in experimental battery cycling data. This algorithm is used for pre-cleaning the experimental cycling datasets for laboratory-fabricated lithium-sulfur (Li-S) batteries, thereby enabling the higher-fidelity fitting of a neural network model to these datasets. The third contribution is a novel algorithm for optimizing the cycling of a lithium iron phosphate (LFP) to maximize the parameter identifiability of a hybrid model of this battery. This algorithm succeeds in improving the resulting model’s Fisher identifiability significantly in simulation. The final contribution focuses on the application of such test trajectory optimization to the experimental cycling of commercial LFP cells. This work shows that test trajectory optimization is s effective not just at improving parameter identifiability, but also at probing and uncovering higher-order battery dynamics not incorporated in the initial baseline model. Collectively, all four of these contributions show the degree to which the selection of battery cycling datasets and experiments for richness and cleanness can enable higher-fidelity data-driven and hybrid modeling, for multiple battery chemistries.