137091 research outputs found
Sort by
Data for Water Ice Formed by Vapor Deposition and Liquid Aerosol Injection: A Comparison Study Using Reflectance Absorption Infrared Spectroscopy
No full textMicrosoft excel files for all data-containing figures (F2-7, S1)Data files for all data-containing figures in: C. E. Buffo, B. M. Jones, and T. M. Orlando, “Water ice formed by vapor deposition and liquid aerosol injection: A comparison study using reflectance absorption infrared spectroscopy,” J. Chem. Phys., vol. 162, no. 24, p. 244509, Jun. 2025, doi: 10.1063/5.0261149.NASA Solar System Exploration Research Virtual Institute (SSERVI) under cooperative agreement number NH22ZDA020C (CLEVER). Grant number: 80NSSC23M022
Understanding the Disuse of Conversational Agents Among Older Adults
No full textSmart technologies have the potential to enhance the aging experience for older adults by helping them age in place and by improving their overall quality of life. However, despite the potential benefits of adopting these technologies, a substantial number of older adults discontinue using them. To gain a deeper understanding of the concept of technology disuse and the underlying factors that contribute to it, a comprehensive literature review
was conducted. This review shed light on the potential roles of impulsivity, psychosocial experience, and user experience with technology in influencing disuse behavior. A mixed-methods research study was conducted with 82 older adults who had previously adopted conversational agent technology but no longer use the device. Results revealed a significant positive correlation between psychosocial experience and user experience, demonstrating the impact of social and environmental factors on older adults’ perceptions of technology usability. Furthermore, the study found no evidence supporting the role of impulsivity in technology disuse. Ways of further exploring this issue are proposed. Lastly, the results indicated that as participants’ psychosocial experience with the technology improved, they exhibited a decreased inclination to abandon it. The qualitative portion of the research unveiled a perceived risk associated with using this technology due to concerns about data breaches and distrust of the companies that made the device. Ultimately, this risk perception also led to disuse among some participants. Additionally, using technology was found to impact participants’ sense of autonomy, which also contributed to disuse. The theoretical and applied contributions of this work are discussed. In conclusion, this research provides a multifaceted understanding of technology disuse among older adults, combining theoretical and practical insights to guide the development of technology that enhances older adults’ experiences.M.S.Psycholog
Development of electrochemical methods for the oxidation and reduction of halogenated organic compounds
No full textHalogenated organic compounds (HOCs) pose a great threat to human health and the aquatic environment because of their environmental persistence and toxicity from carbon-halogen bonds. Electrochemical methods, including electrochemical oxidation (EO) and electrochemical reduction (ER) processes, are promising technologies for the destruction of HOCs because of their advantages of strong oxidation/reduction ability, high efficiency, mild reaction conditions, and high automation. This motivates us to explore and advance the potential for both EO and ER to destroy typical HOCs. Our study aimed to (I) provide effective catalyst design strategies, and (II) bring both mechanistic and kinetic insights for employing both EO and ER processes for the treatment of typical HOCs.
EO has shown effectiveness on the destruction of many HOCs including perfluorooctanoic acid (C7F15COOH, PFOA), an emerging HOC that are ubiquitous in the aquatic environment. However, the mechanisms and preferred pathway for PFOA mineralization in EO remain unknown. In addition, the substantial generation of chlorinated byproducts (i.e., ClOx) during EO also hinders its practical applicability. To overcome these challenges, we made several improvements for EO PFOA treatment including mechanism exploration, kinetic model development, and increasing anode hydrophobicity.
Specifically, in chapter 2, we proposed a plausible optimum PFOA mineralization pathway to determine reaction limiting factors by combining density functional theory (DFT) simulations and experiments. By systematically exploring the roles of valence band holes (h+), hydroxyl radicals (HO•), and H2O, we found that h+ dominated PFOA mineralization. Our proposed primary pathway included oxidation by h+, decarboxylation, and radical fragmentation reactions. Accordingly, PFOA loses COF2 group cyclically without generating short chain perfluoroalkyl carboxylic acids (PFCAs) until complete mineralization.
Besides, in chapter 3, we developed a simplified kinetic model for PFOA degradation in EO and created an energy estimator to predict the energy use for various operational parameters. We considered all the main processes (i.e., mass transfer with electromigration included, surface adsorption/desorption, and oxidation on the anode surface) to quantitatively evaluate the effects of various operational parameters (i.e., applied current densities, initial PFOA concentrations, and flow velocities) on EO PFOA treatment. The applied current density was recognized as the major factor that determines energy consumption, and the predicted PFOA oxidation kinetics and energy consumption by our model agreed with observed experimental results.
Additionally, in chapter 4, the generation of chlorinated byproducts was successfully suppressed by increasing anode hydrophobicity. We fabricated a series of anodes with different hydrophobicity by incorporating various amounts of PTFE on anode surfaces using either electrochemical deposition or membrane casting methods. The resulting PTFE-modified anodes exhibited considerably reduced generation of chlorinated byproducts with the increase of hydrophobicity. In addition, PFOA removal and defluorination efficiencies were also improved.
In addition to EO, ER is also an effective method for the dehalogenation and destruction of HOCs. However, ER is limited by the difficulty of breaking C-F bonds (i.e., one of the strongest carbon bonds) even for the benchmark Pd electrocatalyst. Targeting at eliminating the rate-limiting C-F cleavage step, in chapter 5, we incorporated boron into Pd electrocatalysts, which exhibited remarkably enhanced defluorination and detoxification efficiency. Boron inclusion boosted H* generation (i.e., hydrogen radical that is adsorbed to the catalyst surface) and achieved significantly improved degradation and defluorination efficiency for the model HOC, florfenicol (FLO). DFT simulation and experiments both revealed that boron modification provides a distinct C-F cleavage mechanism that does not involve H*, in addition to the predominantly adopted H*-induced hydrodefluorination mechanism. Boron modification induces excess electrons on surface Pd atoms, which enlarges C-F bond length, shortens Pd-F distance, and turns the thermodynamically unfavorable direct C-F cleavage reactions into favorable reactions.Ph.D.Environmental Engineerin
Demand Projection and Complex Resource Allocation Decisions on Networks
No full textIn this thesis, we consider various practical and theoretical problems arising from health and humanitarian systems that involve the projection of demand and the allocation of resources to populations, which are represented in a network structure. These problems include: (i) quantifying the impact of non-pharmaceutical interventions (NPIs) on disease spread and the need for hospital capacity, (ii) assessing the trade-offs between the public health benefits, such as the reduction in infection spread and adverse outcomes, and other consequences of NPIs, particularly in terms of the number of homebound individuals or person-days, (iii) partitioning a disaster-impacted area among multiple heterogeneous resources to obtain connected components of similar weight, where the weights are node- and resource-specific. To address these problems effectively, we employ mathematical modeling approaches, in particular, agent-based simulation models and optimization-based techniques tailored for each problem, allowing us to capture the network component in our analyses.
Chapter 1 provides a brief introduction and a background to the research problems addressed in this thesis. In Chapter 2, we evaluate the effectiveness and impact of non-pharmaceutical intervention decisions, including school closures, shelter-in-place orders, and voluntary quarantine, using an agent-based simulation model focused on the state of Georgia. We test various scenarios with different durations of shelter-in-place measures and time-varying compliance levels to voluntary quarantine. The outcomes of these simulations provide valuable insights to decision-makers, enabling them to make informed recommendations for social distancing measures to the public. Chapter 3 quantifies the impact of non-pharmaceutical interventions and investigates the trade-offs between their potential benefits, such as reduction in infection spread and adverse outcomes, and socioeconomic consequences of refraining from workplace and community interactions. We measure these trade-offs by considering the number of homebound individuals or person-days and the number of infections and deaths. These evaluations can assist local and national decision-makers in choosing different combinations of interventions over time to reduce infection spread while considering the societal and economic impact.
In Chapter 4, we focus on the problem of partitioning a graph among multiple heterogeneous resources to obtain connected components of similar weight, where the weights are node- and resource-specific. We formulate this problem as novel integer and mixed-integer programs, develop approximation algorithms with provable bounds for special graph and weight structures, and propose several heuristics. We also present an extensive computational study, including a realistic hurricane scenario for the state of Florida, focusing on the post-disaster debris collection application of this problem, where the objective is to move debris from the disaster-affected area efficiently. Chapter 5 focuses on special cases of the problem by considering the connectivity and planarity of the underlying graph, as it significantly affects the problem's computational complexity. We analyze the complexity of various special cases and present approximation algorithms for planar and non-planar graphs with certain connectivity. Building on the theoretical foundations, we introduce heuristics for general graphs. We present the results of a computational study focusing on a hurricane scenario for the state of Florida.Ph.D.Operations Researc
Accelerated Tensor Robust Algorithms for Hyperspectral Imaging and Video Processing
No full textIn recent years, the application of tensor-based methods to high-dimensional data has gained considerable attention, particularly for tasks involving denoising, classification, and compression of complex data structures such as hyperspectral images. This thesis presents novel approaches to enhance Tensor Robust Principal Component Analysis (TRPCA), addressing challenges such as computational efficiency, noise removal, and real-time processing.
Firstly, the thesis presents a new online robust principal component analysis (RPCA) algorithm that recursively decomposes incoming data into low-rank and sparse components. Unlike traditional approaches that operate on data vectors, this method preserves the multi-dimensional structure of data, such as video frames. It is based on the recently proposed tensor singular value decomposition (T-SVD) and incorporates a convex optimization-based approach for recovering the sparse component and updating the low-rank component using incremental T-SVD. An efficient tensor convolutional extension to the Fast Iterative Shrinkage Thresholding Algorithm (FISTA) is also proposed, significantly speeding up the optimization process. The effectiveness of this online tensor-RPCA is demonstrated through its application in background-foreground separation in video streams, where the foreground is modeled as a sparse signal and the background as a gradually changing low-rank subspace. Extensive experiments on real-world videos showcase the robustness and effectiveness of the proposed algorithm.
Secondly, the thesis proposes a randomized blocked algorithm for tensor singular-value thresholding (T-SVT), aimed at reducing the computational demands of TRPCA when applied to noisy hyperspectral images. TRPCA has been successfully employed to reduce noise by employing a minimization involving a tensor nuclear norm and a -norm to separate the low-rank hyperspectral image from the sparse noise. However, the high computational complexity of TRPCA is primarily due to the implementation of the T-SVT operator, which typically involves performing full tensor singular value decomposition (T-SVD) followed by shrinking the singular values of the frontal slices in the frequency domain. The proposed randomized blocked algorithm incrementally finds the singular values until they fall below the threshold, leveraging compression achieved by the fast Fourier transform (FFT) to accelerate TRPCA significantly. Numerical experiments indicate that this method is much faster than traditional TRPCA approaches while maintaining classification accuracy.
Finally, the tensor-robust CUR (TRCUR) method is introduced for hyperspectral data compression and denoising. This method heavily downsamples the input hyperspectral image to form small subtensors and performs TRPCA on these subtensors. The desired hyperspectral image is recovered by combining the low-rank solution of the subtensors using tensor CUR reconstruction. We provide theoretical guarantees showing that the desired low-rank tensor can be exactly recovered using our proposed TRCUR method. Numerical experiments demonstrate that our method is up to 14 times faster than performing TRPCA on the original input data, while maintaining the classification accuracy.Ph.D.Electrical and Computer Engineerin
Sensing Touch from Vision for Humans and Robots
No full textTo affect their environment, humans and robots use their hands and grippers to push, pick up, and manipulate the world around them. At the core of this interaction is physical contact which determines the underlying mechanics of the grasp. While contact is useful in understanding manipulation, it is difficult to measure. In this thesis, we explore methods to estimate contact between humans, robots, and objects using easy-to-collect imagery. First, we demonstrate a method which leverages subtle visual changes to infer the pressure between a human hand and surface using RGB images. We initially explore this work in a constrained laboratory setting, but also develop a weakly-supervised data collection technique to estimate hand pressure in less constrained settings. A parallel approach allows us to estimate the pressure and force that soft robotic grippers apply to their environments, allowing for precise closed-loop control of a robot. Finally, we develop a joint pose and contact estimator which may generalize to internet-scale images. Our model leverages multiple heterogeneously labeled datasets and images with contact labeled by human annotators. Overall, this thesis makes progress towards understanding human and robot manipulation from only visual sensing.Ph.D.Robotic
Democratizing Human-Centered AI with Visual Explanation and Interactive Guidance
No full textArtificial intelligence (AI) systems have been increasingly integrated into our everyday lives, yet how they make predictions often remains obscure to both their developers and the people they impact. The opacity of AI models contributes to their perception as "mysterious"—rendering both developers and those impacted by these models powerless when it comes to aligning AI models with their values. To address these challenges, my research applies a human-centered approach to explain AI models and empower different stakeholders to align AI models with their knowledge and values. This thesis focuses on three complementary thrusts.
(1) Explain AI to Everyone. We pioneer easy-to-access interactive visualization systems that help AI novices and experts understand AI models (e.g., WizMap and CNN Explainer used by 360k+ novices worldwide). We also present first-of-its-kind resources (e.g., 6.5TB DiffusionDB with 14 million prompt-image pairs) to help AI developers and policymakers understand the impacts of large generative AI models.
(2) Guide AI with Human Values. We introduce GAM Changer (deployed by Microsoft) to empower AI developers to vet and fix problematic model behaviors, and GAM Coach to enable those impacted by AI to receive customizable suggestions to alter unfavorable AI decisions.
(3) Democratize Human-Centered AI. We show how researchers can lower the barrier to adopting human-centered AI practices by integrating these practices into AI practitioners' workflows. We highlight an example: Farsight leverages in-situ interfaces to foster responsible AI awareness during AI prototyping.
Our work is making significant impacts on academia, industry, and society. CNN Explainer has been integrated into deep learning courses across top universities, such as CMU, Georgia Tech, Duke University, and the University of Tokyo, receiving over 7k stars on GitHub. DiffusionDB has received over 2M data requests through the HuggingFace APIs. Furthermore, our work has been recognized with 4 best paper type awards at top conferences like ACL, CHI, and FAccT. Additionally, this thesis has been acknowledged by Apple and J.P. Morgan AI PhD Fellowships.Ph.D.Machine Learnin
Adaptive User Interfaces for Personalized Services
No full textPresented on February 6, 2025 at 12:30 p.m. in the Tech Square Research Building, Ballroom.Pat Langley is the principal research scientist in the Information
and Communications Laboratory at Georgia Tech Research Institute and director of the Institute for the Study of Learning and Expertise.Runtime: 59:36 minutesThe Internet has made available massive amounts of information and given users more choices than ever before, but all too often the result is more confusion than satisfaction. Intelligent assistants can help people filter relevant information and guide their choices, but users have different goals and distinctive tastes. In this talk, I report work adaptive user interfaces -- interactive systems that automatically personalize their content to individual users. These incorporate technology and principles from machine learning, intelligent agents, and human-computer interaction to improve the user's experience. I describe a number of prototype systems, including a personalized navigation aide, an adaptive news reader, and a conversational destination advisor. Along the way, I consider design decisions about the problem formulation, the representation of user profiles, the unobtrusive collection of user feedback, and the effective utilization of inferred profiles. I claim that progress on personalized services depends not on development of new algorithms, but rather on the integration of existing methods in novel ways. This talk describes joint work with Nicolas Fiechter, Melinda Gervasio, Wayne Iba, Mehmet Goker, Seth Rogers, and Cynthia Thompson