137091 research outputs found
Sort by
The Past in the Future
No full textInterview portion of Lost in the Stacks, episode 627. Features interview with Dr Dillon Henry, Digital Accessioning Archivist, and Cliff Landis, Digital Curation Archivist, both of the Georgia Tech Archives. They discuss the current status of the ongoing service and space known as "retroTECH" , where students and faculty can engage in hands-on experiential learning of vintage technological devices.Interview portion of Lost in the Stacks, episode 627. Features interview with Dr Dillon Henry, Digital Accessioning Archivist, and Cliff Landis, Digital Curation Archivist, both of the Georgia Tech Archives. They discuss the current status of the ongoing service and space known as "retroTECH" , where students and faculty can engage in hands-on experiential learning of vintage technological devices
Investigating AAVE in Question Answering Systems
No full textThe advancement of technology and social inclusion have encouraged the growth of written texts in dialects. Unfortunately, due to the lack of text corpora, most of the state-of-the-art NLP models are trained by Standard American English (SAE) only. It is important to build NLP technology that is both effective and inclusive. Hence, we investigate the performance of state-of-the-art QA systems on AAVE texts. The performance is examined by converting SQuAD and CoQA to AAVE. Our experiments show that the performance of QA systems degrades significantly when tested on AAVE data in a zero-show setting. While the performance can be partially recovered by incorporating AAVE data in the training set, it still leaves much space for improvement.UndergraduateComputer Scienc
Trustworthy Binary Classifications in Dynamic Systems Under Uncertainty
No full textBinary classifications are commonly used in modeling dynamic systems and are prevalent in machine learning and deep learning applications. These classifications directly inform decision-making, however, uncertainty in modeling can undermine prediction trustworthiness. This dissertation aims to enhance binary classification reliability in dynamic systems under various uncertainties, focusing on safety-critical applications like malware detection and wildfire prediction. In these contexts, predictions inform critical mitigation decisions, necessitating high-fidelity predictions across a range of behaviors that dynamic systems are likely to exhibit. To achieve this, we identify different sources of uncertainty in the modeling pipeline and propose mitigative measures.
In malware detection, we perform online uncertainty estimation to monitor predictions post-deployment, recognizing the model's limitations in classifying unknown behavior as benign or malicious. Next, we develop strategies to faithfully capture malware behavior, reducing or managing uncertainty that manifests as an overlap between the positive and negative classes. Finally, shifting the context to a complex dynamic system like wildfire evolution, we address irreducible uncertainty arising from the system's inherent randomness. Here, we propose a novel evaluation criterion that treats uncertainty as a feature rather than a bug. This criterion tests the model's ability to capture macro system behaviors and does not penalize the model if it accurately learns the system's inherent uncertainty.
We consolidate our learnings into an uncertainty-based framework which can be used for designing reliable data-driven models that perform binary classifications in uncertain dynamic environments. Our findings apply to alert generation using time series and spatiotemporal data collected from sensors monitoring dynamic system behavior.Ph.D.Electrical and Computer Engineerin
Enhancing Sustainable Practices in Secondary Aluminum Manufacturing
No full textIn the recent decades, sustainability has emerged as a crucial concern for
governments and industries worldwide. While the political sphere appears to take the lead
in the pursuit of a greener world from a societal standpoint, industries play a pivotal role
in implementing innovative technologies. However, certain industries still heavily rely on
natural gas consumption for their operations.
Metal manufacturers, in particular, face significant energy demands due to their
melting processes, which have historically necessitated the use of fossil fuels. While the
steel industry has started embracing electrification in its production, the secondary
aluminum industry faces challenges in implementing this technology due to ongoing
research and differences in material properties. Additionally, the aluminum industry has
relied on fossil fuels since its early stages, highlighting the challenges and risks of a full
transition to other energy sources.
While other industries push the electrification movement, a full-scale transition of
the aluminum sector would require large capital investments from the manufacturers.
These financial constraints hinder the industry’s organic progression toward more
sustainable operations. Consequently, the aluminum industry, specifically the secondary
sector, is compelled to enhance its current processes in the race towards sustainability until
new technologies become viable. One prominent industrial-scale approach to address
carbon footprint is the recovery of waste heat. This method seeks to capture excess heat
generated during various industrial processes and utilize it to improve energy efficiency,
thereby reducing greenhouse emissions.
This master’s thesis focuses on examining the current workflow of secondary
aluminum manufacturing, with a specific emphasis on rolled sheet production. By
analyzing combustion intensive equipment involved in the process, it aims to evaluate
machine efficiencies and estimate the recoverable energy present in the flue gasses.
Additionally, the primary focus of this thesis is to identify inefficiencies and energy losses
throughout the production process. These internal inefficiencies are driven by thermal
energy losses caused by internal supply chain. Hence, throughout this paper, the adoption
of just-in-time manufacturing is implemented to analyze the effects of proposed strategies
that can effectively reduce the industry’s global energy demand and carbon footprint.M.S.Mechanical Engineerin
Ghidorah Comes to Visit
No full textA misunderstood three-headed monster from venus just wants to make friends in Tokyo
Spring 2025 Ph.D. Dissertation Lightning Talks
No full textSchool of Interactive Computing PhD students will present 2025 Dissertation Lighting Talks.Presented on April 10, 2025 in the TSRB Building, 1st floor ballroom at 12:00 p.m.Adam Coscia is a fifth-year PhD student in Human-Centered Computing advised by Alex Endert, whose research explores how visual analytics tools can help experts leverage and evaluate AI like LLMs for complex, data-driven tasks in critical domains including education, intelligence analysis, data science, and autonomous exploration of space and deep-sea environments. He engages in user-centered design with analysts, scientists, and engineers to create visual analytics interfaces that not only explain opaque AI behaviors, but also enhance workflow efficiency and domain experts' understanding of the data, revealing novel design implications for bridging cutting-edge AI, visual analytics, and interface design with practical, real-world applications.Yao Dou is a fourth year PhD student at Georgia Tech advised by Prof. Wei Xu. His current research interests lie in natural language processing (NLP), especially text generation, evaluation and online user privacy.Sichen Jin is a fourth-year CS Ph.D. student at Georgia Tech specializing in GIS, visual analytics, and human-computer interaction. Her research focuses on developing a visual analytic tool for exploring how underlying geography affects social ties.Jin Yu is a fifth-year PhD student in human-centered computing, focusing on co-design, educational technology, and interactive systems to empower youth as active creators of future technologies. Her work bridges physical computing and digital platforms, developing toolkits that support hands-on learning, prototyping, and collaborative design.Qiao Zhang is a fifth year PhD student in Computer Science advised by Dr. MacLellan, whose research lies in the intersection of Artificial Intelligence and Human Computer Interaction. Using multi-player cooperative games as environments, Qiao develop agents that can interact with human to understand human-AI teaming dynamics from the perspectives of communication, collaboration and adaptation.Jonathan Zheng is a second year PhD student at the Georgia Institute of Technology majoring in computer science. His current research interests lie in natural language processing, specifically in the learning and reasoning process of large language models.Runtime: 54:11 minutes(Coscia) Title: Visual Analytics for Trustworthy LLMs in Education(Dou) Title: Building Interactive and Personalized AI with User Simulators(Jin) Title: Visual Analytics for Spatial Social Network Mapping and Analysis(Yu) Title: Designing With, Not Just For: How In-Depth User Research Transforms Interaction Design - This dissertation explores how middle school students can be empowered as co-designers of future technologies through hands-on, iterative, and inclusive design processes. By developing and studying three design toolkits—My:Talkies, iReal, and Tangible-MakeCode—this work demonstrates how accessible tools and guided frameworks can help youth engage in ideation, prototyping, and evaluation of interactive systems. Through participatory workshops and educational interventions, the research highlights how deeper engagement in design fosters both learning and more inclusive innovation. The findings contribute to human-centered design by positioning youth not only as users but as creators and evaluators of future technologies.(Zhang) Title: Designing and Evaluating Human-AI Teaming Dynamics in Gaming Environments - To investigate Human-Machine Teaming (HMT) dynamics---and how AI capabilities influence team behavior and performance---we propose a series of three studies to explore communication, coordination, and adaptation in HMT paradigms. To support these investigations, we are developing multiple AI agents and using collaborative games as testing environments to evaluate the human-AI team performance. This work contributes to two central topics in HMT research: 1) the bidirectional adjustments that human and AI agents may develop when working as a team and, 2) how different types of AI agents and interventions can impact the teaming efficiency in human-AI teaming.(Zheng) Title: Investigating LLM Generalizability in Privac
President's Newsletter
No full textWhenever I give a presentation about Georgia Tech, I remind everyone that we have one shareholder only — the people of the state of Georgia — and we need to ask ourselves how everything we do delivers value to them. I am proud to report that the return on investment we provide our state has never been greater, and every summer, I enjoy traveling across Georgia to see it for myself
Fluid simulation on flow maps
No full textIn computer graphics, traditional fluid simulation methods often fail to accurately compute fluid advection, leading to the loss of vortex details and poor visual quality. To address this issue, this work improves a novel flow map method with strong vortex-preserving capabilities [1, 2], extends its application scope[3, 4, 5], and enhances its computational
efficiency[6, 7]. Additionally, the flow map method is applied to simulate various phenomena, including fluid-solid coupling [4], laden particles[5], compressible fluid [8], and free
surfaces [3], achieving state-of-the-art (SOTA) results.M.S.Computer Scienc