UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Scalable Continual Learning Using Cascading Hypernetworks and Cellular Automata
No full textAddressing the critical challenges of catastrophic forgetting and scalability is paramount for developing continual learning systems capable of robust operation in real-world environments. This dissertation presents a novel framework that achieves state-of-the-art performance by integrating two core innovations: a cascading hypernetwork architecture for adaptive model generation and a specialized mini-cellular automata (mini-CA) system for automatic task identification.
The proposed cascading hypernetworks employ a hierarchical structure, where parent hypernetworks can generate the weights of other hypernetworks and eventually task networks, enabling efficient learning across an expanding task set. This is coupled with an auto-generative replay mechanism that reconstructs network samples and replays them to the same hypernetwork, significantly mitigating catastrophic forgetting without reliance on large memory buffers.
For precise task identification, we introduce mini-CA, an efficient and scalable cellular automata variant tailored for complex spatio-temporal data. By utilizing smaller, specialized units with enhanced parallelism and input scalability, mini-CA surpasses the limitations of traditional CA. The system is adeptly extended to process colored images and sequences and integrates seamlessly with deep learning architectures for multi-channel input.
In the unified framework, mini-CA first identifies the incoming task, which then directs the cascading hypernetwork to dynamically generate a tailored, task-specific network. Rigorous evaluations on reinforcement learning and image classification benchmarks demonstrate superior accuracy, latency, and scalability. This research underscores the transformative potential of combining hierarchical hypernetworks with specialized CA to build truly adaptive and continuously learning systems for complex applications.Electrical and Computer Engineerin
Novel Methodologies for Spatial-Temporal Dependent Data
No full textThe application of spatial-temporal data across various fields such as environmental science, public health, urban planning, building management, transportation, agriculture, and more. Weather data is typical spatial-temporal data that often plays a critical role as an input feature for many environmental-related files. However, the spatial-temporal dependency in the data presents a significant challenge, especially in modeling high-dimensional dependent data. This complexity arises due to the intricate spatial and temporal dependencies that influence the accuracy and efficiency of statistical models. Motivated by the unique challenges of high-dimensional data, this dissertation focuses on innovative methodologies for predicting and analyzing weather data, which has inherently spatial-temporal dependency in nature.
The first chapter presents an innovative approach by utilizing the crowd-sourcing weather data from personal weather stations to improve the prediction accuracy using a general spatial-temporal modeling framework by ensemble local forecasts of the target location made using neighboring personal weather stations. Our approach distinguishes from existing literature in three key ways: firstly, it leverages crowd-sourced weather data from personal weather stations alongside traditional public data sources, enhancing both temporal (e.g., 5-minute intervals) and spatial resolutions (specific locations versus grids). Secondly, it utilizes spatial-temporal correlations between the target and neighboring weather stations, allowing for a nuanced capture of complex underlying patterns. Finally, our model is adaptable for making forecasts for buildings located at any arbitrary location. We equip the proposed modeling framework with random forest as a representative of advanced machine learning models. Using data collected from San Antonio, Texas as a case study, we analyze the impacts of model parameters, to gain insights on how many neighboring personal weather stations are needed to guarantee the forecast accuracy and how strong the temporal dependence on the input weather features is to forecast the future temperature. Using a building located at arbitrary location in San Antonio, we demonstrate the advantages of the proposed framework by comparing with forecasts using airport temperature and persistence model. We also study the performance of the model at different hours of a day and at different seasons.
The second chapter is motivated by the weather data that were collected for the first chapter but shifts focus to multivariate analysis. In this chapter, we propose a spatial-temporal partial envelope (ST-PENV) model that is parsimonious and effective in modeling high-dimensional spatial-temporal data. The partial envelope model is proposed under a linear coregionalization model framework which allows heterogenous spatial-temporal covariance structure for different variables of the response vector. The maximum likelihood estimator for the proposed model can be obtained through a Grassmann manifold optimization. We obtain an asymptotic result for the estimator and conduct a thorough simulation study to demonstrate the soundness and effectiveness of the proposed method. We also apply the proposed model to analyze the crowdsourcing weather data collected from personal weather stations in the city of Syracuse, NY of the United States.
In the third project, we extend our exploration by proposing a Bayesian framework to the spatial envelope model. By integrating prior information of model parameters, offering a more flexible and robust framework for capturing the uncertainty and spatial correlation in high-dimensional weather data. We investigate appropriate prior specifications, study the propriety of the posterior distribution, and implement posterior sampling. A simulation study is conducted to evaluate the model’s performance using spatial dependent data.
Overall, this dissertation develops and demonstrates novel methods to address the significant challenges of dependent data analysis. These methodologies not only improve the accuracy of weather forecasts but also have broader applications in various scientific and practical fields.Management Science and Statistic
OPPORTUNITY YOUTH'S JOURNEY: EXPLORING EARLY-LIFE FACTORS AND HEALTH IMPLICATIONS IN TRANSITION TO ADULTHOOD IN THE UNITED STATES
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until May 15, 2027.This study examines the early-life factors associated with opportunity youth status and the resulting health implications during the transition to adulthood in the United States. Specifically, this study examined; 1) the influence of early-life socioeconomic status and neighborhood disadvantage on opportunity youth status in the U.S; 2) the impact of opportunity youth status on young adult self-rated health and depressive symptoms in the U.S; and 3) the influence of past opportunity youth status on early midlife self-rated health and depressive symptoms. This study utilized the restricted National Longitudinal Study of Adolescent to Adult Health (Add Health), Waves I, III-V. The findings suggest that about 1 out of every 8 young adults (12%) are disconnected from education and employment—opportunity youth—in the U.S. Additionally, individuals from low socioeconomic backgrounds and high disadvantaged neighborhoods exhibit higher chances of being an opportunity youth. The findings show that opportunity youth are more likely to report poor/fair self-rated health and higher levels of depressive symptoms compared to other youth. More so, individuals who were once opportunity youth are more likely to experience poor/fair self-rated health and high levels of depressive symptoms in early midlife. Lastly, the findings from this study highlight the need to tackle factors hindering upward mobility and health disparities for young adults who are disconnected from school and work.Demograph
Clash of Civilizations? An Analysis Deconstructing the Binaries in Antony and Cleopatra
No full textThis thesis explores to what extent Shakespeare’s play Antony and Cleopatra reinforced stereotypes regarding European superiority and inter-racial character dynamics, and how these racial dynamics influenced gender dynamics. While some argue that his works perpetuate certain stereotypes prevalent in his time, others contend that Shakespeare’s portrayals were more nuanced and complex, allowing characters more agency and gravitas rather than being reduced to stereotypes.
Antony and Cleopatra present the clash of civilizations as a prominent theme in this play, representing binary oppositions when it comes to racial and gender identity constructions—notions that break down in the play. In addition to reading the play through these concepts, I explore how receptions, performances, and renditions of the play in more modern times grappled with the ripple effect of Renaissance racial constructions. My goal was to test both the validity and limitations of the clash of civilizations thesis as applied to Antony and Cleopatra, both through a close analysis of the play and through a review of its production and reception history. Shakespeare, through Cleopatra’s character, was not only depicting Renaissance ideas but also subverting them simultaneously. The binary of the cultural divisions can be both exaggerated and dismantled; the play, while it does have outdated concepts of race and ethnicity, is ideologically challenging in a way that continues to confront internal biases even in recent times.Englis
The Latina Superintendent in Rural Texas Districts: The Challenges and the Successes
No full textThis qualitative dissertation aimed to understand Latina women's lived experiences as they ascended to and served as school district superintendents. The literature review confirmed the oppressive forces faced by Latina leaders and highlighted the lack of mentorship programs available to them. It also revealed the cultural wealth and skills that Latina leaders bring to top leadership positions.
To capture the essence of their experiences, five participants who served as district superintendents in K-12 non-charter public school districts in South Texas took part in this study. The study employed pláticas and testimonios to represent the lived experiences of Latina superintendents effectively. These cultural approaches were used to amplify participants’ voices and experiences. Using a Chicana Feminist Epistemology, the data was analyzed to highlight the opportunities and challenges faced by the participants.
Findings from this study emphasize the need for a comprehensive examination of systems that support and promote Latina leaders into district leadership roles. It calls upon policymakers, evaluators of leadership programs, and state and national leaders to implement necessary changes to increase the number of Latina superintendents in Texas and across the nation.Educational Leadership and Policy Studie
¿Por Que se Translada la Genta?; Todos Somos Tejanos; Los Tejanos de Hoy
No full textThe Texans One and All curriculum at the Institute of Texan Cultures examines the migration, settlement patterns, cultural traditions, and lasting contributions of diverse ethnic groups to Texas. The curricula in this series was originally intended to supplement a guided tour through the comprehensive exhibits featuring over 20 cultural communities—including Native American, African American, European, Asian, and Middle Eastern groups— which documents the multicultural foundations of Texas. These curricula provide insights into how various peoples have collectively shaped the state's identity and heritage
Developing Military Service-Related PTSD Training Curricula for Entry Level Counselors: A Delphi Study
No full textThe demand for both mental health counselor representation in Veterans Health Administration facilities and support from community providers underscores the need for enhanced counselor training with military populations. A literature review showed a lack of guidelines addressing appropriate counselor training and supervision to treat military service-related posttraumatic stress disorder (PTSD) issues. To address the call for research into the preparation of counseling students to work with PTSD-diagnosed military populations, this dissertation sought to develop a training curriculum consensus for entry-level clinical mental health counselors to work with PTSD-diagnosed service members and veterans. This study uses a Delphi research method to explore the experience of military counseling experts. This Delphi study consisted of three rounds, with up to 13 expert panelists answering each round. The guiding research questions informed the Round One questionnaire, asking expert panelists about curriculum components, placement of components in the CACREP curriculum, internship experiences, and supervision. The two following rounds utilized curriculum component statements developed from Round One responses. Participants rated the appropriateness of each of the developed curriculum component statements. A total of 106 statements met consensus through three complete Delphi rounds. Further discussion addresses the results and implications for counseling programs, counselor educators, and supervisors. This study concludes with limitations and suggestions for future research related to training entry-level counselors to work with PTSD-diagnosed service members and veterans.Counselin
Modeling Spiny Projection Neuron Development in Tourette's Using iPSC Organoids
No full textTourette Syndrome (TS) is a complex neuropsychiatric disorder characterized by involuntary motor and vocal tics. It has been associated with dysfunction in the cortico-striatal circuit in the brain, but the exact cause is still not fully understood. Immunocytochemical analyses have shown a reduction of GABAergic and cholinergic interneurons in the striatal postmortem tissue of adults with TS when compared to controls. However, the development of the spiny projection neurons that create the circuit have not been investigated. To investigate how their development might contribute to the dysfunction happening in the striatum of TS patients, we generated striatal organoids using human induced pluripotent stem cells from a TS patient and a healthy family member as a control. Analysis of the spiny projection neuron population within the TS-derived organoids revealed a significant reduction in SPNs along early development. Using an assembloid method, we were able to observe how cortical neurons project into the striatal organoids using our TS-derived organoids, mimicking the circuit found in the brain. Further understanding of how the deficit of SPNs in the striatum affects the circuit will potentially help to better understand TS and its development.Neuroscience, Developmental and Regenerative Biolog
Between Curry and Kale: Eating Disorder Risk Among Adult Women of Indian Descent Living in the United States
No full textIndian Americans, individuals of Indian descent living in the United States, remain underrepresented in counseling research, particularly in eating disorder (ED) studies. Given that eating disorders comprise the second most fatal subset of mental illness and present differently across cultures, understanding their risk factors in Indian American women is crucial. However, research of this population is lacking. The purpose of this study was to explore the symptoms and characteristics, as identified by the Eating Attitudes Test-26 (EAT-26), of EDs prevalent among women of Indian descent living in the United States while examining predictive variables through self-report surveys to understand which variables contributed to ED risk in this population. Specifically, I hypothesized that dieting symptoms would be more common than oral control or bulimia symptoms. I also hypothesized that age, immigrant status, generational status, religious commitment, and self-objectification would predict eating disorder risk. To test this, I conducted a quantitative study using self-report surveys including the EAT-26, the Religious Commitment Inventory-10, and the Self-Objectification Scale. I ran t-tests and multiple regressions to test my hypotheses. Findings indicated that dieting subscale symptoms of the EAT-26 are the most prevalent in this population. Additionally, religious commitment and self-objectification were significant predictors of eating disorder risk. This is the first study to explore eating disorder risk among Indian American women, highlighting a critical gap in research. Implications for counseling, counselor education, and supervision are discussed, along with recommendations for future research to further explore complexities of eating disorder risk in this population.Counselin
Addressing Inconsistencies Across Software Environments
No full textSoftware environments are vital to modern software development, offering tools and frameworks that enhance productivity, reduce costs, and improve software quality. Consistency across different environments is crucial for reliable and efficient builds, tests, and executions. Variations in hardware, operating systems, software dependencies, and configurations can lead to significant challenges and inefficiencies. Although approaches like containerization, virtualization, continuous integration/continuous deployment (CI/CD), and infrastructure as code (IaC) strive to make environments more reproducible and consistent, issues related to environments are still common, particularly for those with additional restrictions or limited domain knowledge.
This dissertation addresses two significant research problems related to inconsistencies across software environments. Firstly, we introduce PExReport, a novel framework designed to extract cross-project failures (CPFs) along with their execution environment. PExReport automates the generation of stand-alone executable CPF reports by leveraging build systems to prune source code and dependencies, creating a pruned environment that accurately reproduces CPFs. Demonstrated within the Maven build system as PExReport-Maven, our framework achieved a high reproduction rate of 92.93%, successfully reproducing 184 out of 198 CPFs with exact failure types and messages. Additionally, PExReport-Maven significantly reduced the required source classes and dependencies, enhancing the efficiency of CPF reproduction. Secondly, we explore the challenges faced by non-contributors, specifically Computer Science students, in handling build issues in open-source software (OSS). Our pilot study collected data from command history logs, environment variables, network information, and virtual machine snapshots, revealing that non-contributors often struggle with certain build issues. Furthermore, a dual-phase study involving 330 build tasks among 55 students characterized non-contributors' build issues and resolution attempts. An intervention in Phase II successfully improved the build success rate by proactively addressing difficult "trap" issues.
Overall, this dissertation introduces the PExReport, which addresses environmental inconsistencies in CPF reproduction. Additionally, it presents comprehensive studies aimed at resolving build environment inconsistencies from the perspective of non-contributors. This dissertation offers tools and methodologies to mitigate these inconsistencies across various software environments and identifies potential directions for future research.Computer Scienc