UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Leveraging Latent Fields for Accurately Attributing Model Behavior
No full textLarge Machine Learning and Artificial Intelligence models have undeniably become essential tools for addressing many modern problems, so the need for explainability and transparency with respect to such models is now greater than ever. Many existing methods for computing attributions lack a strong underlying explanation for their utility, and are inflexible with respect to differing user requirements. Furthermore, many existing approaches to problems in computer vision fail to efficiently utilize the information present in the data, or fail to take advantage of opportunities for a problem-centric design of solutions. This work proposes several approaches increasing the explainability of existing models, and for approaching new problems from a perspective of transparent design. The primary contribution of this work is a novel formulation of integrated attributions for generating informative statistics with respect to model inputs and model parameters. These Generalized Integrated Attributions provide a transparent means of extracting diverse sources of information regarding high-dimensional input and parameter spaces, resulting in improved interpretability as well as increased utility for applications such as strategic training and unlearning. Additionally, this work describes methods for increasing data efficiency in model training schemes, and identifies several opportunities for explainable design in addressing common computer vision problems. All causal explanations are inherently subjective, and no tool will ever be guaranteed to perform perfectly as intended, but by internalizing the principles of explainability, transparency, and interpretability, we can more quickly and efficiently develop statistics and models which are more useful and reliable.Computer Scienc
Air Quality in the Built Environment: Investigating Pollutant Exposure, Ventilation and Artificial Intelligence Methods
No full textThis dissertation investigates the impact of various ventilation strategies on indoor air quality (IAQ) in unique building typologies that are underrepresented in the literature. The first study examined a naturally ventilated cross-training gym, finding that natural ventilation was insufficient. CO2 and particulate matter levels frequently exceeded NIOSH standards, particularly on colder days when ventilation was reduced. The study concluded that tailored ventilation methods are necessary to maintain optimal IAQ in such facilities. In the second study, IAQ data from the cross-training facility was used to develop an Artificial Intelligence (AI) tool designed to predict optimal occupancy intervals for maintaining IAQ. Specifically, a Long Short-Term Memory (LSTM) neural network model was developed to predict these intervals, demonstrating the potential to enhance occupant health and safety by identifying safe time slots. The model effectively designated times with lower pollutant levels, proving superior to human decision-making. This approach underscores the potential of predictive analytics in managing IAQ in indoor environments. The third study evaluated the use of a heat recovery ventilation (HRV) system in a tiny house. The results revealed the HRV's effectiveness in reducing particulate matter and volatile organic compounds compared to natural ventilation, despite variations in effectiveness depending on the pollutant and activity. The study highlighted the importance of mechanical ventilation strategies in managing IAQ in compact dwellings. Collectively, these findings provide comprehensive insights into the effectiveness of different ventilation methods and tools, emphasizing the need for customized strategies to ensure healthier indoor environments across various unique and under-researched building typologies.Architectur
Development of Polypeptide Vaccines Against Acinetobacter Baumannii
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 September 11, 2027.Multi-drug resistant (MDR) Acinetobacter baumannii is an opportunistic bacterial pathogen associated with hospital-acquired infections. This pathogen’s extensive MDR phenotype offers limited treatment options and requires alternative immunotherapeutic interventions, such as prophylactic vaccination in susceptible populations. In this study, we employed immunoinformatics to identify peptides containing both putative B and T cell epitopes from proteins associated with A. baumannii pathogenesis. An Acinetobacter Multi-Epitope Vaccine (AMEV2) comprising an A. baumannii thioredoxin A leader protein followed by 5 identified peptide antigens was created. Mice sensitized to a sublethal challenge dose of the pathogen generate antibodies that recognize AMEV2's epitopes. Subcutaneous immunization of mice with AMEV2 plus AddaS03 adjuvant demonstrates a robust humoral immune response, resulting in high antibody titers to the antigen and its peptide components. Immunization generates a Th2 response with an increased frequency of IL-4 and IL-5 secreting T cells following restimulation with the whole protein and its peptides. Immunized mice were protected from a lethal intraperitoneal and intranasal challenge with hypervirulent strains of A. baumannii and achieved 80% and 60% survival rates, respectively. AMEV2 vaccination provides antibody-mediated protection against hypervirulent A. baumannii infections, with passive vaccination of naive mice with AMEV2 antisera affording 67% protection. In vitro opsonophagocytosis assays demonstrate that anti-AMEV2 serum enhances the killing of opsonized bacteria by macrophages. This study demonstrates the creation of a newly designed vaccine that stimulates antibody-based protective immune responses for mice against A. baumannii infection. These data suggest a novel therapeutic strategy against an extremely drug-resistant pathogen with few remaining treatment options.Molecular Microbiology and Immunolog
Unlocking Academic Success: Exploring Associations Between 24-Hour Movement Compositions and Academic Performance in College Students
Get PDF• Higher academic achievement in college can facilitate more job opportunities (French et al., 2015) and plays a significant role in the hiring process (Rynes et al., 1997)
• 24h movement behaviors (i.e., sleep, physical activity, sedentary behavior) have each been independently linked with academic performance (Felez-Nobrega et al., 2018; Okano et al., 2019; Taylor et al., 2013; Wald et al., 2014)
• Mixed results with physical activity and sedentary behavior
• Sleep is positively associated with academic performance
• Only one study has examined 24h movement behaviors as a collective with academic performance (Pellerine et al., 2023)
• No studies to date have employed compositional data analysis techniques which limits bias in estimates by considering the codependence of the behaviors across a whole day (Dumuid et al., 2018)
Purpose: Examine how the composition of time spent engaging in 24h movement behaviors relates to academic performance among college studentsPsycholog
Robust Techniques to Detect and Mitigate Volumetric and Non-Volumetric Network Attacks
Get PDFWith the quick evolution of the computer and communication networks, there is an increase in the network attacks. The growing number of attacks, make network security an important and a pressing problem. The network attack (malicious traffic) detection and mitigation, a major part of network security, has been a widely researched topic for decades. Various approaches, from simple filtering to complex deep packet inspection, were proposed and evaluated. Recently, the research focus shifted towards the application of artificial intelligence (AI), including machine learning (ML) and deep learning (DL), to network security with promising results. The features representing network traffic patterns play a vital role in AI-based network traffic analysis or attack detection methods. However, the currently used features are based on the statistical data of network traffic, in aggregate or individual flows, and yield mixed results. AI models trained and tested on a network traffic dataset perform well; when tested for the same attacks on datasets collected from a different network environment, the attack detection accuracy can be as low as 50%. The primary reason for such low accuracies is the distribution of the features of test data being different from those of the trained data. This is considered an out-of-distribution (OOD) problem. This research focuses on the analysis of volumetric (a large number of packets or flows/second) and non-volumetric (lower rate but exploit protocol features extensively) network attacks and their mitigation techniques. We investigate the effectiveness and limitations of the attack-specific mitigation techniques, the application of AI techniques for network attack detection, and the statistical features. We construct behavioral attack-specific features (ASFs) to enhance the detection of the same or similar attacks whenever and wherever they occur without being susceptible to the OOD problem while reducing the computational overhead and detection lags. We propose a framework to automate the construction of the ASFs using large language models (LLMs) demonstrate its implementation.Computer Scienc
Making the American Gorilla: Exploring The Production of Endangered Species and Conservation in America
Get PDFFrom King Kong to Koko, gorillas hold a unique place in American imaginations and are intimately tied to our perceptions of nature, conservation, and relationships with the environment. Since their first introduction to Western science by American naturalists in 1848, Americans have encountered gorillas in a variety of contexts, including film, comics, news articles, and natural history museums. However, the most visible and visceral encounters occur within zoological institutions, where the North American gorilla population is (re)produced and managed across 50 institutions for public consumption. In these diverse contexts, gorillas are embedded within specific cultural and conservation ideologies, which have changed dramatically over time. In large part, these shifts have been driven by the environmental and conservation movements, first influenced by the work of conservationists like Dian Fossey and led today by a network of gorilla experts working across zoological institutions and their in situ conservation partners. Gorillas, as a captive species and wild animal, emerge from and co-produce this close-knit professional network, made legible to Americans in the virtual and physical encounters with gorillas mediated by these professionals. Using a mixed-methods approach, this dissertation traces the role of American gorilla agency in shifting American perceptions of the species and how endangered species and global conservation networks co-produce species conservation and the American conservation ethic in a time of conservation crisis.Anthropolog
School Closures in the Neighborhood: A Case Study in One School District
No full textThere is a sense of melancholy whenever a neighborhood school closes its doors forever. Communities lose identity with the loss of a beloved and familial, societal hub. Schools are where academic dreams and goals are born. The joy of a new school year and the social circle of friends in a thriving community give students the fresh start for reaching their full potential. Our democracy requires that every child is entitled to a free and public education regardless of their zip code. As cities shrink and the urban core loses population to suburban migration, school districts are faced with empty buildings and fewer students (Bierbaum, 2010).
Fewer students in expensive buildings causes less opportunities for students to achieve the American dream. After the pandemic, returning to in-person instruction became a challenge. Urban schools were faced with cutting costs and one solution was closing or consolidating campuses. Urban gentrification or ?shrinking cities? pushed school districts to ?right size? aging, older facilities with fewer fund dollars (Bierbaum, 2020, p. 450-452).
The entire community surrounding the school is affected. Successful schools build communities with thriving businesses, health and social services. Effective schools influence the ?geography of opportunity? for urban cities (Tate, 2008, p. 397). Neighborhoods change identity when a school closes. There is a loss of community identity when the heartbeat of a neighborhood school closes its doors for the last time.
My case study research explored the lived experiences of families and educators before, during and after the school closures. I chose this research design to explore the effects of school closures on families. I served as the main researcher and I sought individuals who experienced school closure. My research explored the experiences and decisions that families faced during school closure and the effect of school closure in their neighborhoods.Educational Leadership and Policy Studie
A Review of AI-Based Cyber-Attack Detection and Mitigation in Microgrids
No full textIn this paper, the application and future vision of Artificial Intelligence (AI)-based techniques in microgrids are presented from a cyber-security perspective of physical devices and communication networks. The vulnerabilities of microgrids are investigated under a variety of cyber-attacks targeting sensor measurements, control signals, and information sharing. With the inclusion of communication networks and smart metering devices, the attack surface has increased in microgrids, making them vulnerable to various cyber-attacks. The negative impact of such attacks may render the microgrids out-of-service, and the attacks may propagate throughout the network due to the absence of efficient mitigation approaches. AI-based techniques are being employed to tackle such data-driven cyber-attacks due to their exceptional pattern recognition and learning capabilities. AI-based methods for cyber-attack detection and mitigation that address the cyber-attacks in microgrids are summarized. A case study is presented showing the performance of AI-based cyber-attack mitigation in a distributed cooperative control-based AC microgrid. Finally, future potential research directions are provided that include the application of transfer learning and explainable AI techniques to increase the trust of AI-based models in the microgrid domain.Electrical and Computer Engineerin
Characterization of Trabecular Bone Microarchitecture and Mechanical Properties Using Bone Surface Curvature Distributions
Get PDFUnderstanding bone surface curvatures is crucial for the advancement of bone material design, as these curvatures play a significant role in the mechanical behavior and functionality of bone structures. Previous studies have demonstrated that bone surface curvature distributions could be used to characterize bone geometry and have been proposed as key parameters for biomimetic microstructure design and optimization. However, understanding of how bone surface curvature distributions correlate with bone microstructure and mechanical properties remains limited. This study hypothesized that bone surface curvature distributions could be used to predict the microstructure as well as mechanical properties of trabecular bone. To test the hypothesis, a convolutional neural network (CNN) model was trained and validated to predict the histomorphometric parameters (e.g., BV/TV, BS, Tb . Th, DA, Conn.D, and SMI), geometric parameters (e.g., plate area PA, plate thickness PT, rod length RL, rod diameter RD, plate-to-plate nearest neighbor distance NND<sub>PP</sub>, rod-to-rod nearest neighbor distance NND<sub>RR</sub>, plate number PN, and rod number RN), as well as the apparent stiffness tensor of trabecular bone using various bone surface curvature distributions, including maximum principal curvature distribution, minimum principal curvature distribution, Gaussian curvature distribution, and mean curvature distribution. The results showed that the surface curvature distribution-based deep learning model achieved high fidelity in predicting the major histomorphometric parameters and geometric parameters as well as the stiffness tenor of trabecular bone, thus supporting the hypothesis of this study. The findings of this study underscore the importance of incorporating bone surface curvature analysis in the design of synthetic bone materials and implants.Mechanical EngineeringBiomedical Engineerin
Regulating Modality Utilization within Multimodal Fusion Networks
Get PDFMultimodal fusion networks play a pivotal role in leveraging diverse sources of information for enhanced machine learning applications in aerial imagery. However, current approaches often suffer from a bias towards certain modalities, diminishing the potential benefits of multimodal data. This paper addresses this issue by proposing a novel modality utilization-based training method for multimodal fusion networks. The method aims to guide the network&rsquo;s utilization on its input modalities, ensuring a balanced integration of complementary information streams, effectively mitigating the overutilization of dominant modalities. The method is validated on multimodal aerial imagery classification and image segmentation tasks, effectively maintaining modality utilization within <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mo>&plusmn;</mo><mn>10</mn><mo>%</mo></mrow></semantics></math></inline-formula> of the user-defined target utilization and demonstrating the versatility and efficacy of the proposed method across various applications. Furthermore, the study explores the robustness of the fusion networks against noise in input modalities, a crucial aspect in real-world scenarios. The method showcases better noise robustness by maintaining performance amidst environmental changes affecting different aerial imagery sensing modalities. The network trained with 75.0% EO utilization achieves significantly better accuracy (81.4%) in noisy conditions (noise variance = 0.12) compared to traditional training methods with 99.59% EO utilization (73.7%). Additionally, it maintains an average accuracy of 85.0% across different noise levels, outperforming the traditional method&rsquo;s average accuracy of 81.9%. Overall, the proposed approach presents a significant step towards harnessing the full potential of multimodal data fusion in diverse machine learning applications such as robotics, healthcare, satellite imagery, and defense applications.Electrical and Computer EngineeringComputer Scienc