Theses and Dissertations

Item

A pipeline for data and knowledge extraction from material science literature to accelerate scientific discovery

(Georgia Institute of Technology, 2023-08-16) Shetty, Pranav

Scientific literature is growing at an exponential pace which makes it difficult for scientists to search through and effectively utilize the data contained in it. In this work, we developed methods and data sets needed to extract knowledge and material property data from a corpus of 2.4 million materials science articles. We uniquely identified extracted polymer materials by training supervised clustering models using parameterized cosine distances with hierarchical agglomerative clustering that achieve state-of-the-art results on a benchmark data set of polymer named entity clusters. In addition, we built sequence labeling models that can tag property information using an ontology specific to the materials domain. MaterialsBERT, a pre-trained encoder fine-tuned on the aforementioned corpus of materials science papers was used as the encoder for the sequence labeling model and outperforms the baselines tested for data sets in the materials domain. We developed two pipelines, one that combines sequence labeling outputs with heuristic rules, and another using prompts to a large language model, to extract material property records from our corpus of papers. The extracted data is made available to the public through the interface polymerscholar.org. A subset of the extracted data was used to train machine learning models to predict the power conversion efficiency of polymer solar cells, thus demonstrating an end-to-end pipeline that goes from literature-extracted data to data-driven insights. This work will reduce the time taken during the search as well as the discovery phase of experimental work, thus allowing researchers to move beyond an Edisonian trial-and-error approach.

Item

Visual shape and pose recovery for robotic manipulation

(Georgia Institute of Technology, 2024-08-01) Lin, Yunzhi

The objective of this thesis is to extract visual shape and pose information for robotics manipulation tasks, which would enhance the robot’s capability to operate in less-structured scenes. Manipulation is a multi-step task consisting of sequential actions applied to an object, including perception, path planning, and closure of the gripper, followed by a task-relevant motion with the grasped object. Perception, which serves as the primary input source, has been widely studied. Recent research has primarily focused on the task of grasping objects effectively. The goal is to enable robotic systems to grasp a diverse range of objects with high accuracy. While human grasping capabilities at a young age may seem effortless, precise robot grasping remains a formidable challenge due to the vast diversity of objects a robotic arm could grasp and the intricate contact dynamics associated with specific robot hand designs. Deep learning has emerged as a promising approach to address these challenges by detecting SE(2) × R² grasp representations associated with parallel plate grippers. However, existing methods still suffer from two related issues: sparse grasp annotations or insufficiently rich data, leading to covariate shift problems. Another limitation lies in the grasp configurations themselves. Most algorithms are designed for the bin-picking problem, which is task-agnostic and only performs top-down grasps. However, specific scenarios may impose constraints or limitations, necessitating the exploration of better solutions that permit various task-relevant grasp configurations. It is worth noting that shape information presents an alternative approach to tackling the problem aforementioned, as explicit shape representations can inform fine-grained grasping operations. Beyond how to grasp, how to manipulate is a more demanding task. It requires the robot to be aware of target-centric information. This ability includes locating objects and their poses, also known as the 6-DoF pose estimation problem (i.e., 6 degrees of freedom, from 3D position + orientation). Accurate, real-time pose information of nearby objects in the scene would allow robots to engage in semantic interaction. The problem of pose estimation is a rich topic in the computer vision community, yet most existing methods have focused on instance-level object pose estimation. Such methods suffer from a lack of scalability. On the other hand, category-level object pose estimation opens the door to work on all the targets within a specific category, which promises to scale better for real-world applications. More recently, generalized object pose estimation has attracted more attention as it removes the assumption of known instances or categories. It is more applicable and accessible compared to the methods mentioned above. In this thesis, we focus on improving robotic manipulation with primitive shape recognition, category-level pose estimation and tracking, generalized pose estimation and tracking, and multi-level robotics scene understanding. A series of methods are proposed to improve the applicability of real-world robotic manipulation. To alleviate the data insufficiency problem and generate multiple 3D grasp configurations, we propose a segmentation-based architecture for decomposing objects sensed with a depth camera into multiple primitive shapes, along with a post-processing pipeline for robotic grasping. Segmentation employs a deep network, trained on synthetic data with 6 classes of primitive shapes and generated using a simulation engine. Each primitive shape is designed with parametrized grasp families, permitting the pipeline to identify multiple grasp candidates per shape region. The grasps are rank-ordered, with the first feasible one chosen for execution. For category-level object pose estimation, we propose a simple and efficient RGB-based approach (without depth) that only requires oriented 3D bounding box annotations at training time and thus does not require CAD models for training. This design decision allows us to take advantage of large collections of real-world images. We also extend the design to the tracking problem, which further incorporates uncertainty estimation through a tracklet-conditioned deep network and a filtering process. To further expand the scalability of the object pose estimation methods, we investigate the inverse use of parallel NeRF for robust object pose estimation in a render-and-compare manner. It can be applied to any novel object, removing the limitation of known category assumptions. We also explore the generalized object pose tracking problem in dynamic environments. We develop a streamlined pipeline combining video segmentation, uncertainty-aware keypoint refinement, and structure from motion, effectively tracking 6-DoF poses from short-term monocular RGB video. We also generate a large-scale photo-realistic synthetic dataset for training and evaluation. Finally, we establish a comprehensive scene representation for advanced manipulation, which includes high-fidelity 3D reconstruction, a rough approximation of primitive shapes, and accurate object pose estimation. Throughout this research, we study how to extract visual shape and pose information for real-world manipulation scenarios. The rest of the thesis is organized as follows. Chapter 1 covers a high-level idea of the related fields. Chapter 2 focuses on shape recognition for object grasping. Chapter 3 introduces a keypoint-based RGB-only category-level pose estimator. Chapter 4 extends it into a tracker via a tracklet-conditioned network and filtering process. Chapter 5 explores the idea of inverse use of NeRF for pose estimation in parallel. Chapter 6 explores the large-scale dataset and uncertainty-aware keypoint estimation for object pose tracking. In the end, Chapter 7 introduces a multi-level scene representation with multi-view RGB inputs for the robotics manipulation task.

Item

Empirical Measurements of the Security, Privacy, and Usability of Website Password Authentication Workflows

(Georgia Institute of Technology, 2024-07-31) Alroomi, Suood

In an era where digital interactions are integral to daily life, the security and privacy of online authentication mechanisms are crucial for protecting user data and maintaining trust in web services. Passwords, though decades old, remain the most common form of authentication and are likely to stay ubiquitous. Therefore, the web ecosystem’s security depends on how users and websites handle passwords and manage authentication. Researchers have extensively explored user behavior with passwords, offering insights into how websites should handle authentication and leading to significant updates in modern guidelines. A significant gap remains in understanding how websites handle authentication and whether they adhere to best practices. This dissertation aims to bridge that gap through large-scale empirical measurements of website authentication practices. I develop measurement techniques to systematically evaluate websites’ authentication policies and implementation decisions and apply them at scale to assess their authentication workflows. I reveal the disparity between modern recommendations and real-world implementations. My studies show that while guidelines inform policy decisions, barriers prevent adopting recent recommendations, highlighting the need for education and outreach efforts. Further, I found poor policy decisions aligning with the default configurations of web software, which often compromise security, privacy, or usability. Updating these defaults to match modern guidelines could significantly reduce vulnerabilities and promote best practices. Moreover, incorporating security features such as blocking common passwords and rate limiting could significantly enhance the security of websites, as many are found lacking these defenses. I also identify concerning practices in authentication workflows, such as insecure communication, misconfigured HTTPS deployments, and mixed content vulnerabilities. While TLS deployment has improved, work remains to migrate all sensitive resources to HTTPS. Standardized authentication workflows with centralized security controls and outreach efforts can further mitigate inconsistencies and improve authentication security.

Item

Promoting And Deactivating Effects of Carbonaceous Deposits During Skeletal 1-Butene Isomerization Over Ferrierite

(Georgia Institute of Technology, 2024-07-30) Hebisch, Karoline L.

Most microporous solid acid catalysts deactivate during hydrocarbon conversion because of carbon depositions (“coke”) blocking pores and poisoning active centers. However, several cases of reaction enhancement have been reported. One reaction in which coke has a promoting effect is the skeletal isomerization of linear butene to iso-butene over the zeolite ZSM-35 (framework type FER, possessing perpendicular intersecting 8-R and 10-R channels). While consensus exists about the reaction mechanism during reaction startup, the reaction location and the reaction pathway at peak catalyst performance are still contested. Time-resolved catalyst characterization data are collected to (1) understand the effects of high temperature (T=420 °C) and carbon deposition on the zeolite micropore structure and Brønsted acid site accessibility and (2) the location and chemical nature of carbon deposits. Three distinct reaction stages are identified: catalyst startup (0-24 h), optimal performance (50-300 h) and catalyst deactivation (>300 h). Most of the deposits (~5 wt%) form within the initial 24 h and are located inside the micropores, rendering them effectively inaccessible to probe molecules (e.g., N2 and Ar) and leading to an expansion of the crystal unit cell. Adsorption isotherms of several hydrocarbons revealed that only small molecules with a kinetic diameter of <4.7 Å can diffuse into the pores before they become obstructed by carbon deposits. Calculation of diffusion coefficients from transient adsorption data at reaction temperature shows that even small molecules are severely hindered in their diffusion, concluding that the reaction occurs at the pore entrances. Operando reactivity quenching with basic probe molecules with different steric constraints shows that acid centers exist under reaction conditions for tens of hours, are located in the pore mouths, and can be reversibly poisoned. Because a combination of steric confinement, acidity, and carbonaceous deposits is needed to successfully facilitate skeletal isomerization, the active sites are concluded to be monoaromatic species, which form during catalyst startup in the pore mouths and are protonated by internal Brønsted acid sites. The positive charge is delocalized and communicated via a methyl group to the catalyst exterior, where skeletal butene isomerization is facilitated. The outstanding activity of ferrierite for this reaction is explained by FER’s ability to anchor the catalytically active deposits in the pores while preventing their premature deactivation by hindering side-chain growth and condensation reactions. Catalyst deactivation is explained by the formation of external, polyaromatic condensates preventing the reactants from accessing the active sites in the pore mouths. With this information, the catalyst performance is optimized by selective oxidation of residual organic structure directing agent (OSDA). Three promoting effects of residual OSDA are identified. Residual OSDA selectively poisons the strongest Brønsted acid sites, thereby substantially suppressing side product formation while also improving catalyst lifetime. Carbonaceous fragments further act as a precursor for the active site, thereby shortening the unselective startup phase. Lastly, two catalyst regeneration strategies are explored – oxidation and supercritical fluid extraction. Characterization data of oxidatively regenerated samples shows that coke combusts in a manner similar to a shrinking-core model. If done under milder conditions (500 °C in air flow) for 30 min – 1 h, a substantial amount of deactivating species can be removed while the internal deposits remain largely intact. In contrast, due to higher solubility and increased mobility of small aliphatic, olefinic, and monoaromatic species, internal deposits are preferentially removed during solvent extraction with supercritical CO2. Based on preliminary performance and catalyst characterization data, recommendations for future regeneration approaches are derived. Understanding the activating and deactivating effects on catalyst activity is crucial for prolonging catalyst lifetimes and increasing the efficiency of regeneration processes, which will help the transition from a fossil-resource-based economy to a bioeconomy.

Item

Graphs and geometry: an interplay between local and global views

(Georgia Institute of Technology, 2024-07-27) Yu, Jing

In this dissertation, we explore problems related to graphs and geometry. This work consists of two projects, and they are independent and utilize distinct proof techniques. However, they share a common underlying philosophy: we alternate between local and global perspectives as required. In Project I, we investigate the large-scale geometry of Borel graphs of polynomial growth. Krauthgamer and Lee showed that every connected graph of polynomial growth admits an injective contraction mapping to $(\mathbb{Z}^n, \|\cdot\|_\infty)$ for some $n \in \mathbb{N}$. We strengthen and generalize this result in a number of ways. In particular, answering a question of Papasoglu, we construct coarse embeddings from graphs of polynomial growth to $\mathbb{Z}^n$. Furthermore, we extend these results to Borel graphs. Namely, we show that graphs generated by free Borel actions of $\mathbb{Z}^n$ are in a certain sense universal for the class of Borel graphs of polynomial growth. This provides a general method for extending results about $\mathbb{Z}^n$-actions to all Borel graphs of polynomial growth. For example, an immediate consequence of our main result is that all Borel graphs of polynomial growth are hyperfinite, which answers a well-known question in the area. Additionally, our results yield nice applications in graph minor theory. In Project II, we investigate outerplanar graphs with positive Lin–Lu–Yau curvature. we show that all simple outerplanar graphs with minimum degree at least 2 and positive Lin-Lu-Yau curvature on every edge have maximum degree at most 9. Furthermore, if G is maximally outerplanar, then G has at most 10 vertices. Both upper bounds are sharp.

Item

From microbes to whale sharks: how studying some of the smallest and largest organisms can inform elasmobranch biology and ecology

(Georgia Institute of Technology, 2024-07-29) Perry, Cameron

Whale sharks (Rhincodon typus) are the largest extant fish in the sea; however, there are still large knowledge gaps in their biology and ecology. Information about whale shark reproduction and mating has proven difficult due to logistical constraints of studying a large highly migratory pelagic species. Small oceanic islands can provide insight into these environments and ex-situ field research identified a unique and reliable aggregation of whale sharks in waters surrounding the remote South Atlantic Island of St. Helena. Sharks arrived seasonally in St. Helena waters from December to May each year, peaking in January. Using photo-ID, a total of 277 individual sharks were identified, consisting of a 1.1:1 sex ratio of male and female sharks ranging from 5-12 meters in total length, with 86% of males and 51% of females estimated to be mature. Eyewitness accounts of mating behavior were reported by reliable local observers on two separate occasions, which comprise the first observations of copulation in this species and are consistent with the size and sex demographics of the population. Acoustic telemetry showed that animals use the habitats around the entire island but are focused on the leeward side. Horizontal movements away from the island proved difficult to track, due to deep-diving behavior that either damaged or caused premature detachment of the archival satellite tags, however, some individuals showed large scale movement away from the island towards both Africa and South America. Deployment of CATS camera tags and MiniPAT tags allowed for exploration of the subsurface/diving behaviors and conspecific interactions of whale sharks in St. Helena. Deep dives (>100 meters) in St. Helena were dynamic, characterized by steep pitch angles and activity at depth. However, deep dives were not uniform suggesting that individual context and motivation are important for the function of the dives. Some evidence supports that deep dives may be associated with searching/foraging behaviors due to changes in activity and observed behaviors at depth. However, the cause of these behaviors is unknown and whether they are linked to reproductive behaviors cannot be ruled out. Pre-copulatory and social behaviors were observed on video further supporting that St. Helena is a unique location for whale shark reproductive ecology and conspecific interactions. Due to its likely role in the reproductive ecology of the whale shark, St. Helena represents a critical habitat for this endangered species. Elasmobranchs (sharks, skates and rays) are of broad ecological, economic, and societal value. These globally important fishes are experiencing sharp population declines as a result of human activity in the oceans. Research to understand elasmobranch ecology and conservation is critical and has begun to explore the role of body-associated microbiomes in shaping elasmobranch health. There have been burgeoning efforts to understand elasmobranch microbiomes, exploring microbiome variation among gastrointestinal, oral, skin, and blood-associated niches. I identified major bacterial lineages in the microbiome, challenges to the field, key unanswered questions, and avenues for future work. There is a need to prioritize research to determine how microbiomes interact mechanistically with the unique physiology of elasmobranchs, potentially identifying roles in host immunity, disease, nutrition, and waste processing. Understanding elasmobranch–microbiome interactions may be critical for predicting how sharks and rays respond to a changing ocean and for managing healthy populations in managed care. Elasmobranchs are exposed to a plethora of microbes as they move through their aquatic environment in both horizontal and vertical dimensions. Therefore, understanding of environmental microbiomes is important for complete understanding of host-associated microbiomes in both natural and artificial settings. Longitudinal sampling of a large public elasmobranch aquarium exhibit revealed a dynamic and evolving water column microbiome across 18 months. Water column microbial dynamics were driven by a diverse nitrifying community likely linked to the initial seeding of the exhibit and waste products from the inhabitants. Few microbial members were observed across all sampling time points suggesting that there are significant selection processes that alter community composition, such as ozonation and treatment of the water. Addition of sharks and fish into the exhibit had small but measurable effects on community dynamics and composition. A final “stable” state was not reached over the course of sampling suggesting that microbial succession dynamics may still be occurring or that stable final community states may not be a reasonable expectation in managed care environments.

Item

Inverse Hybrid Bonding with Metal Organic Framework as Infill for Heterogeneous Integration

(Georgia Institute of Technology, 2024-07-28) Sahay, Rohan

This thesis presents Inverse Hybrid Bonding (IHB) as a novel fine-pitch die-to-die (D2D)/die-to-wafer (D2W) bonding scheme for heterogeneous integration. IHB is a two-step process wherein (1) Direct copper bonding is used for I/O connections, and (2) ALD-CVD deposition of Metal Oxide Framework (MOF) is used for the infill. The post-bond infill can potentially mitigate extreme particle control requirements for fine-pitch D2D/D2W bonding frameworks. Furthermore, compared to the viscous epoxy-based conventional underfills, a conformal, void-free infill deposition can be achieved with minimal die-to-substrate gap using gaseous precursors/reactants in the ALD-CVD infill deposition process. In the present work, direct Cu-Cu thermocompression bonding with formic acid preclean was first demonstrated at 300°C. The bond quality was characterized by I/O resistance and shear strength measurements. Furthermore, ruthenium was studied as passivation capping over copper to achieve low-temperature Cu-Cu bonding. For the post-bond infill in IHB, two different materials – aluminum oxide and ZiF-8 MOF were analyzed. A sequential set of experiments were conducted to study the effect of ALD process parameters on area coverage for infill, based on which the ALD/CVD infill process was optimized to successfully demonstrate IHB with a 5 mm x 5 mm effective bonded area and a 1 µm die-to-substrate gap. The infilled material was characterized for coverage and the effect of infill on the mechanical strength of the package. Further, current progress and future directions on IHB for fine-pitch have been discussed.

Item

Redefining Trade Dynamics: Inorporating Network Centrality, Spatial Heterogeneity, and Optimal Transport Theory into Models of International Trade

(Georgia Institute of Technology, 2024-07-27) Helfrich, Ian T.

This dissertation explores new approaches to modeling international trade, focusing on network effects, spatial distributions, and optimal resource allocation. The research is structured in three interconnected chapters, each addressing a distinct aspect of trade dynamics. The first chapter investigates how network centrality measures can enhance our understanding of trade patterns. By constructing a trade network using BACI-CEPII data from 1995 to 2015, the study augments traditional gravity models with various centrality measures. The results suggest that a country's position within the global trade network significantly influences its trade relationships, beyond what economic size and distance alone can explain. Chapter two introduces a novel method for measuring effective distance between trading partners. This approach uses nightlights and population data to create weighted centroids, capturing shifts in economic activity and population distribution over time. Applied to U.S. interstate trade in 2017 and global trade flows from 2015 to 2020, this measure shows promise in improving gravity model estimations compared to conventional distance measures. The final chapter develops a theoretical framework that unifies equilibrium theory and optimal transport in infinite-dimensional spaces. This work proves the existence and uniqueness of equilibrium under general conditions and characterizes these equilibria as solutions to optimal transport problems. While theoretical in nature, this approach offers new insights into market efficiency and resource allocation in complex economic systems. Collectively, this research aims to advance our understanding of international trade by incorporating network effects, spatial heterogeneity, and optimal transport theory into existing models. These approaches may offer improved explanatory power and policy relevance compared to traditional methods, though further research is needed to fully assess their impact and applicability.

Item

Electrospun Hydrogel Based Drug Delivery System for Immune Regeneration in a Cleft Palate Mice Model

(Georgia Institute of Technology, 2024-07-27) Iyer, Keerthana Srini

Cleft palate is a developmental birth defect, defined by an abnormal space or gap formed in the upper lip or palate. This condition has high mortality rate if not treated early. Surgical correction is the current practice in the clinical field, however this comes with the risk of a secondary surgery in these patients. Wound healing takes place in a bacteria laden environment, hence increasing complications and hampers tissue regeneration. There have been many studies proving the efficacy of biomaterials as an alternative treatment plan for patients with cleft palate. Building on this finding, we developed a novel electrospun hydrogel material composed of PEGylated norbornene and thiolated hyaluronate. Developed using photo click chemistry, these materials provide an excellent scaffold for tissue regeneration, while doubling as a drug delivery carrier. The tunability of these materials facilitates in the development of a more robust material for the application of cleft palate repair, making it a potent strategy for wound healing.

Item

Causal Discovery from Observational Data in the Presence of Latent Confounders and Other Data Complexities

(Georgia Institute of Technology, 2024-07-27) Yang, Yuqin

Causal discovery aims to recover causal relationships among variables of interest in the system. In the situations where interventions (controlled experiments) on system variables are not allowed, causal discovery from only observational data has been studied, which either utilizes the conditional independence relations among observed variables, or asserts additional semi-parametric assumptions on the underlying model. However, there are complexities in real-life data that make causal discovery even more challenging. Some of the main sources of data complexity include: (i) Latent confounding, where there may exist unobserved variables that affect more than one observed variables in the system; (ii) Deterministic relations, where one observed variable may be fully dependent on other observed variables in the system; (iii) Measurement error, where we may not observe a exact value of the variables, but rather a corrupted version of them; (iv) Data heterogeneity, where the data are collected from multiple domains and do not follow the same distribution. The majority of causal discovery methods assume that these complexities are absent in the system. Naturally, naive applications of these approaches to the settings that indeed are subject to data complexity issues lead to detecting spurious or erroneous causal links among variables of interest. The focus of the dissertation is on developing causal discovery methods that are capable of handling these data complexities. Specifically, -We study the problem of causal discovery in linear causal models with deterministic relations and latent confounding. We provide necessary and sufficient conditions for unique identifiability of the model under separability condition (i.e., the matrix indicating the independent exogenous noise terms pertaining to the observed variables is identifiable). -We study the problem of causal discovery in linear causal models in the presence of latent confounding and/or measurement error. We characterize the extent of identifiability of the model under separability condition together with two versions of faithfulness assumptions. We provide graphical characterization of the models that are observationally equivalent. -We study the problem of learning the unknown intervention targets in linear or nonlinear causal models from a collection of interventional data obtained from multiple environments. We propose LIT algorithm which allows latent confounders to be intervention targets. Our theoretical analysis shows that LIT algorithm gives a more accurate estimate of the intervention target set than previous works.

Theses and Dissertations

Permanent URI for this collection

Search within this collection.

Browse

Georgia Tech Library

Browse

Recent Submissions