UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Tiled Matrix Multiplication Unit for Edge Devices: Architecture, Implementation, and Floorplanning
Tiled Matrix Multiplication (TMM) is an efficient technique for large-scale matrix operations, playing a crucial role in advancing state-of-the-art artificial intelligence (AI) applications on edge devices. TMM enhances computational efficiency while optimizing hardware resource utilization for large matrix computations. Many companies have integrated Tiled Matrix Multiplication Units (TMMUs) into modern advanced processors; however, the hardware design and implementation of these units remain proprietary and undisclosed. This study aims to unveil the architecture, implementation, and floorplanning of a TMMU, providing insights into its potential fabrication feasibility. The work includes Verilog-based hardware design, simulation results, and power estimation analysis, along with discussions on development challenges and future improvements. The proposed TMMU supports two operation modes: basic matrix multiplication and tiled matrix multiplication. Utilizing a customized CISC instruction set and leveraging a previously developed 4×4 Tensor Processing Unit (TPU), with a clock cycle time of 20 ns per cycle, the TMMU efficiently performs a 4×4 matrix multiplication in 3,000 ns (150 clock cycles) and an 8×8 matrix multiplication in 17,000 ns (850 clock cycles). Future optimizations are expected to enhance the design's performance, as this project serves as a foundational study for developing high-performance ASICs.Electrical and Computer Engineerin
Constraining the atmosphere of Europa using the Space Telescope Imaging Spectrograph
Europa’s tenuous atmosphere primary consists of O2, H2, H2O, O and H, with other trace species present as well. Investigating Europa’s ultraviolet auroral emissions can aid in constraining the composition of the atmosphere especially since O and H have emission lines in this range. The Space Telescope Imaging Spectrograph (STIS) on the Hubble Space Telescope (HST) can observe these emissions, along with UV emissions of possible trace species such as S, C, and Cl. HST observations of Europa’s UV aurora, including the STIS data used in this study, have led to some important findings about the moon including evidence for plumes. These emission observations have also been essential in constraining the abundances of O2, H2O, and H at Europa. Over the years, there have been repeat observations of Europa in the far-UV using STIS that have added up to a large amount of total integration time on the target. By compiling all the observed spectra, we can increase the Signal-to-Noise Ratio (SNR) potentially allowing us to place constraints on further species that are harder to detect. To extract the spectrum we use, we start by subtracting the background signal for the entire STIS image in each dataset. We then isolate the rows in the image that correspond to Europa’s disk and the immediate surrounding environment. Combining the spectra of each row, we then subtract the reflected sunlight albedo emitted from the surface in the non-eclipse observations (when Europa is not in Jupiter’s shadow) to arrive at a final extracted spectrum of the aurora. All spectra across all datasets are then combined to one total spectrum that we use for our constraints. We report on our early findings of the derived species abundances, such as H2 and Cl, in Europa’s atmosphere as well as upper limits for those with high uncertainty.Physics and Astronom
Archaeological Report, No. 521
In July of 2024, the Center for Archaeological Research (CAR) at the University of Texas at San Antonio (UTSA) performed archaeological monitoring at Mission Nuestra Señora de la Purísima Concepción de Acuña (41BX12; hereafter referred to as Mission Concepción) in San Antonio, Bexar County, Texas in response to a request from Pugh Constructors, Inc. This project was carried out in support of maintenance at Mission Concepción, which will replace the gravel walking path that currently connects Mission Concepción to the Padre Margil Pilgrimage Center, Pope Francis Center, and the Margie and Bill
Klesse Garden (hereafter referred to as the Pilgrimage Center) with a concrete sidewalk in compliance with requirements and regulations in the Americans with Disabilities Act (ADA). The Area of Potential Effect (APE) covers approximately 196 m2 between the Pilgrimage Center entrance gate and the granary area of Mission Concepción. Because the property is managed by the National Park Service (NPS), this project falls under Section 106 of the National Historic Preservation Act of 1966, as amended. NPS determined that an Archaeological Resources Protection Act (ARPA) permit was not required due to the scale of work and will have final review authority on the project (Kemp 2024). The project also required review by the Texas Historical Commission (THC) due to its status as a State Antiquities Landmark under the Antiquities Code of Texas (Texas Natural Resource Code, Title 9, Chapter 191, and section 191.092). A Texas Antiquities Permit (TAP), No. 31845 was granted to CAR by the THC in July of 2024. Cynthia Munoz, Interim Director of CAR, served as Principal Investigator, and Mikaela Razo served as Project Archaeologist. The project accession number is 2936.
In February of 2024, CAR excavated six preliminary shovel tests under TAP 31576 within the APE in anticipation of the sidewalk construction. All shovel tests were positive for cultural materials dating from the colonial period to the modern era, and one shovel test revealed a wall remnant (Feature 1; Kemp 2024). It was determined that Feature 1 would not be disturbed by construction, so it was covered and backfilled; however, considering the density of cultural materials and the feature located within the APE, CAR recommended archaeological monitoring with matrix screening during the sidewalk excavation.
During the sidewalk excavation in July of 2024, CAR staff located two additional features within the APE: a possible wall footing presenting as a cluster of mortar and rubble, and a brick-and-mortar walkway. Additionally, dozens more artifacts ranging from the colonial period to the modern era were collected from excavated soil. The proposed sidewalk is not expected to have any further impacts on the features; therefore, the CAR recommends that construction be allowed to proceed as planned. All project-related materials, including the artifacts recovered and the reports generated, will be permanently stored at the CAR curation facility. Artifacts will also be entered into the NPS Interior Collections Management System used by NPS to manage the curatorial collections for Mission Concepción.Pugh Constructors, Inc.Center for Archaeological Researc
PATIENT-SPECIFIC INTRALUMINAL THROMBUS FORMATION POTENTIAL AND HEMODYNAMICS MODELING OF ABDOMINAL AORTIC ANEURYSM GROWTH
An abdominal aortic aneurysm (AAA) is a critical vascular condition marked by unpredictable growth and a high rupture mortality rate exceeding 80%. Defined as a localized aortic enlargement (≥ 3 cm), it typically occurs between the renal arteries and aorto-iliac bifurcation. Untreated AAAs can expand by 1 cm annually, risking rupture. Surgery or endovascular intervention is advised at diameters ≥ 5.5 cm in men or 5.0–5.4 cm in women. Hemodynamic forces, crucial for vascular remodeling, are thought to influence AAA development, though their role is complex due to intraluminal thrombus (ILT) present in over 75% of cases. ILT disrupts direct blood-vessel interaction, complicating AAA biomechanics. High-fidelity hemodynamic modeling aims to predict ILT formation, enhancing understanding of AAA progression, risk assessment, and clinical management strategies.Mechanical Engineerin
The Relationship Between Sensory Processing Difficulties and Gait Metrics in Children With Autism Spectrum Disorder
Children with Autism Spectrum Disorder (ASD) often show motor impairments, including delayed development and postural asymmetry, along with atypical sensory responses like hypersensitivity or reduced vestibular processing. These may contribute to distinct gait patterns such as “tippy toe” walking. There is limited research that has examined the direct relationship between sensory processing profiles and gait kinematics in children with ASD. This study examined how sensory processing difficulties affect gait kinematics in children with ASD across different environmental conditions.
Twenty-four children with ASD and nineteen typically developing (TD) peers participated. Kinematic data was collected using inertial measurement unit (IMU) sensors that were placed on the lower limbs of the participants as they walked across a flat, foam, and beam surface. Gait metrics such as stride intervals and joint angles were analyzed using MATLAB, alongside sensory profile scores from the Sensory Processing Measure (SPM).
Results showed no significant group differences in mean joint angles; however, children with ASD demonstrated significantly greater variability in right and left ankle angles under the beam condition. (p = 0.01 for the right ankle, p = 0.01 for the left ankle). Significant stride interval differences were found within both groups across walking surfaces. A negative correlation was found between stride intervals and Balance and Motion scores in the ASD group (r = -0.473, p = 0.035), suggesting that higher sensory difficulties were associated with shorter, more frequent steps. Several additional correlations between joint angles and sensory domains were found in the TD group.
Overall, this study highlights the link between sensory processing difficulties and gait variability. It offers an insight into how these interactions may inform tailored therapeutic approaches.Health, Community and Polic
The Indirect Influence of Psychological Inflexibility and Quality of Life through Post-Traumatic Stress Symptoms
Approximately 5 – 9% of college students endorse PTSD (Elhai et al., 2012; Read et al., 2011), which negatively affects quality of life (QoL; Polizzi et al., 2022). Understanding modifiable factors that influence the relationship between post-traumatic stress symptoms (PTSS) and QoL is critical. A potential modifiable factor is psychological inflexibility (PI) a set of rigid patterns of behavior to control feelings, thoughts, and internal experiences or avoid unwanted experiences (Levin et al., 2014). PI is a risk factor for PTSS and decreases in PI lead to decreases in PTSS severity (Crabtree et al., 2021) and is associated with worse QoL in college students (Arslan, 2024). It is hypothesized that an indirect effect will be observed such that higher PI will be associated with higher PTSS, which in turn, will be associated with lower QoL.Psycholog
Energy Performance and Preservation in Historic Buildings: Evaluation and Optimization of Passive Cooling Strategies in Hot and Humid Climates
As climate change accelerates, optimizing the energy performance of historic buildings is crucial for their sustainable preservation and resilience. Ensuring the continued use of these structures, along with the thermal comfort and well-being of their occupants, requires a nuanced understanding and optimization of their inherent energy characteristics. This dissertation examines low-thermal-mass historic buildings, a prevalent typology in the U.S. and globally, particularly in hot and humid climates, among the most challenging for passive cooling. Through three interconnected studies, this research assesses the energy performance of a wood-frame heritage structure in San Antonio, Texas. It quantifies the building’s passive cooling potential, evaluates various natural ventilation strategies as a viable means of improving energy performance and thermal comfort, and models the resilience of such structures under future climate conditions. The findings demonstrate that historic preservation and energy efficiency can be successfully integrated. The passive cooling capabilities of these buildings’ inherent features, coupled with natural ventilation, significantly enhance indoor thermal conditions while reducing reliance on active cooling systems. However, as climate conditions continue to warm, hybrid strategies that incorporate both passive and active cooling will become increasingly necessary. This research contributes to the broader discourse on sustainable historic preservation, offering insights into optimizing energy performance without compromising architectural integrity. By leveraging inherent energy-efficient features and passive cooling strategies, historic buildings can maintain occupant comfort and delay the need for mechanical intervention, ensuring their viability in a changing climate.Civil and Environmental Engineerin
Automatic Segmentation of Abdominal Aortic Aneurysm from Computed Tomography Angiography Using SAM and XAI
This poster was presented at the 2025 Postdoctoral Appreciation Week event.Segmentation of abdominal aortic aneurysm (AAA) in 2D computed tomography (CT) images is essential for accurate measurement and treatment planning, but general-purpose AI tools struggle with grayscale CT patterns or require manual intervention. We introduce AAA-SAM, a fully automatic deep-learning pipeline that adapts a general segmentation model called the segment anything model (SAM) to become an AAA specialist using adapters. The low-rank adaptation (LoRA) and the convolutional adapter fine-tune the existing vision skills of the model to recognize the unique textures and contrasts of abdominal CT scans. The mask generator produces aneurysm boundary maps at multiple resolution levels. These outlines are binarized and converted into bounding-box prompts that the model uses to generate a detailed, high-resolution aneurysm outline without human intervention. Subsequently, we up-sample the model’s initial outline to full CT resolution and refine it using a combined binary cross-entropy, Dice, and focal loss to improve region overlap and boundary accuracy. Finally, the base model’s behavior on complex and incorrectly predicted samples is analyzed with explainable artificial intelligence (XAI) techniques to provide insights into its inference process, which in turn is utilized to reinforce the design of the model. This end-to-end process runs in seconds per image frame. Using publicly available benchmark CT data sets, it achieves faster and more accurate results than existing segmentation models and conventional supervised methods. The thinking processes of different sub-modules within the model are revealed through XAI during runtime, enabling experts to identify remaining errors and updating the instance without additional training by selectively switching sub-modules on or off. We trained and evaluated AAA-SAM in a distributed data-parallel regime, logging Dice similarity, Intersection over Union, and 95th-percentile Hausdorff distance in Weights and biases. Based on an open data set of 2D CT benchmarks, AAA-SAM consistently outperformed SAM and fully supervised baselines, handling irregular shapes and low contrast boundaries with high speed and accuracy suitable for a real-time clinical decision support system. A comprehensive evaluation across these benchmarks demonstrates the superior performance of AAA-SAM while updating only 0.65% of the SAM model parameters via LoRA.Computer Scienc
Cut to the Core: Development and Validation of a Short Version of the Multidimensional Sense of Belonging-21 (MSBI-21)
The present study aimed to develop and validate a short form of the Multidimensional Sense of Belonging Inventory–21 (MSBI-21), a self-report instrument designed to assess three dimensions of belonging: Negative Emotion (NE), Social Reassurance (SR), and Intrinsic Motivation (IM). Given the length of the original 21-item scale and the need for short, psychometrically sound tools, this study constructed and evaluated a 12-item version (MSBI-12) using two simulated datasets (N = 785 and N = 460) generated based on Palomin’s (2020) thesis. Study 1 examined the factor structure, reliability, and item discrimination parameters of the MSBI-21 using exploratory and confirmatory factor analyses, along with item response theory (IRT) and coefficient omega (ω) estimates. Study 2 tested the performance of the short form using Exploratory Structural Equation Modeling (ESEM), reliability estimates (ωu), and the Quantifying Convergent Validity (QCV) index to evaluate relationships with concurrent constructs including social support, belongingness, and psychological distress. Across both studies, MSBI-12 demonstrated strong internal consistency, structural validity, and adequate convergent evidence. These findings show initial patterns of the MSBI-12 as a reliable and efficient measure of the multidimensional nature of belonging, with potential applications in psychological research and higher education settings. Limitations and directions for future research using real human data are discussed.Psycholog
Application of the Transition State Theory in the Study of the Osmotic Permeabilities of AQP7, AQP10 and GlpF
Aquaglyceroporins, including human AQP7, AQP10, and <i>E. coli</i> GlpF, are known to facilitate movements of glycerol, water, and some other uncharged molecules across the cell membrane. In this study we focused on the transport of water molecules in the absence of glycerol for AQP7, AQP10 and GlpF using the Transition State Theory for the novel application of permeability and kinetics studies. We conducted around 500 ns of in silico simulations of the aquaglyceroporins embedded in lipid bilayer membranes with intracellular-extracellular asymmetries in leaflet lipid compositions. For the water permeability analysis, we computed the transition rate constant with correction for recrossing events where the water molecules do not completely traverse the protein channel from one side of the membrane to the other side. We also studied the hydrogen bond distributions of the single-file waters and channel residues and linear water densities along the pores of the aquaglyceroporins. Interestingly, we found that there was an inverse correlation between the number of single-file water molecules in the channel and osmotic permeability.Physic