116964 research outputs found
Sort by
Scholarly footprints : the impact of students on urban evolution
No full textScholarly Footprints: The Impact of Students on Urban Evolution is a quantitative study that set out to understand how the large student population in Austin, Texas has changed the city from 2000 to 2023. The indicators mapped using ArcGIS Pro include total population, population between the ages of 18 and 24, total housing units, owner-occupied housing units, renter-occupied housing units, and average gross rent. The findings indicate that growth in the total population and total housing units occurred across the city but was the most intense on the edges of the city. Around the Higher Education Institutions included in the study area, there was a general increase in the surrounding tracts in terms of the population aged 18 – 24, renter occupied housing units, and average gross rent. These findings are significant for planners, as they highlight how student populations shape neighborhood dynamics and influence housing demand, affordability, and the spatial distribution of density—insights that are key to planning for equitable and sustainable growth in university-adjacent neighborhoods.Community and Regional Plannin
Comparative evaluation of novel neural network architectures
No full textA neural network is a methodology for function approximation. For a long time, Multi-Layer Perceptron (MLP) has been a primary architecture for neural network-based applications. Recently, several other neural network architectures have been proposed by researchers, each supported by different theorems related to function approximation, such as XNet, Kolmogorov-Arnold Network (KAN), and Fourier Analysis Network (FAN). This report attempts to highlight the characteristics and capabilities of each architecture. This report explores each architecture from the perspective of function approximation. The experiments are carried out using eight univariate functions of different types. Functional accuracy and the variance measurements are presented here. Based on the experimental results, we suggest use cases for each architecture that demonstrate its characteristics and capabilities as follows: XNet might be useful in image recognition, FAN can be useful in time-series can be modelled, KAN might be suitable for scientific experiment analysis, and MLP can work as a wild card that can be used as the baseline of the models listed above.Informatio
Sparklefurs : virtual avatars, physical embodiment, and intergenerational histories in the furry fandom
No full textAlthough optimism about the internet’s liberatory potential declined after the burst of the dot-com bubble, the furry fandom has continued to view computer-mediated communication as a space for identity exploration beyond the limits of the physical body. This thesis examines the paintings and textile works of contemporary furry artist Mark Zubrovich, in which he embodies his sparkledog fursona, Bruce—an avatar derived from digital art subjects of the late ‘00s, characterized by fantastical powers and queer-coded fashion. I situate Zubrovich’s practice within the historical context of the internet’s widespread adoption, drawing on intergenerational dialogues within the furry fandom and scholarly discourse on cyber-utopianism. By identifying Zubrovich and other furry artists as producing historically and affectively engaged works featuring sparklefurs, I assert their significance as participants in visual traditions concerned with nonhuman embodiment and the expression of queer desire.Art Histor
Assessing Communication Gaps Between Health Advocates and Texas Legislators in Efforts to Enhance Public Health Outcomes
No full textThe backbone of a healthy society is its legislation. When we envision a state that prioritizes the health of its citizens, we must enable a body of lawmakers who are not only literate on current health crises, but can form productive, long-lasting relationships with the health advocates that have their fingers on the pulse of a variety of afflicted communities. Unfortunately, Texas legislators continue to underperform in bettering health outcomes and underutilize the research and tactics of countless health professionals, non-profit and advocacy organizations to address health issues. Current literature in health communication acknowledges these missed opportunities but fails to provide actionable solutions. In this work, I analyze the disconnects between lawmakers and health advocates in the context of Texas legislation, and highlight positive outcomes borne out of effective health communication between both parties. These foundational observations inform communication strategies for health advocates to adapt their advocacy approaches, encouraging open dialogue legislators at the local and state level. Findings from this narrative literature review may guide future interventions and educational models to equip health advocates and lawmakers in co-opting legislation that betters health outcomes across Texas.College of Natural Science
Repeat offenders : experienced coup makers, partisan officers, and the coup trap
No full textUsing detailed case evidence from Syria and Iraq, I find that the coup trap, the tendency for countries where a coup has occurred to experience more coups, is driven by the tendency of experienced coup-makers to plan and execute more coups. Officers who take part in one coup acquire the skills and inclination to participate in others. Thus, they pose a special danger to the ruling coalition. It is not easy for leaders to safely remove these officers because, while leaders know these officers are a threat, the officers have supporters and confederates who might rise to their defense, or retaliate against the leader. How easy it is for coalitions to coup proof depends on how likely these experienced coup-makers are to end their coup plotting permanently. When party organizations develop within the military, this coup-maker attrition becomes more likely. A military with party organizations inside it has more identifiable factions, and so in the churn of coups and purges, it is easier for victors to remove the support networks of rival experienced coup-makers and neutralize the threat they pose. Syria and Iraq each went through several rounds of repeat coup-makers before the Ba’ath party consolidated power over the military, allowing for a transition to more stable dictatorships under Hafez al-Asad and Saddam Hussein, respectively.Governmen
From bricks to clicks : consumer preferences in the digital age
No full textThe rapid digitization of society, accelerated by technological advancements and the COVID-19 pandemic, has transformed daily life by shifting interactions from traditional in-person engagements (“bricks”) to digital experiences (“clicks”). This shift has altered how individuals participate in activities, affecting travel demand. The current dissertation investigates this phenomenon by examining behavioral responses to digitization across three key domains: e-commerce, healthcare, and dining. A research framework comprising four independent yet conceptually connected components is developed to evaluate how individual characteristics and built-environment factors influence consumer decisions in digitally mediated settings.
In the e-commerce domain, the dissertation addresses the growing issue of product returns, which poses logistical challenges to supply chains and transportation systems. We investigate how sociodemographic and spatial characteristics influence online shopping frequency, return rates, and return channel choices (e.g., in-store, mail, parcel lockers) to inform strategies that reduce unnecessary travel and promote sustainable logistics.
In the healthcare domain, the research focuses on the digital transformation of service delivery through telemedicine. We examine the timing of adoption relative to the pandemic and investigate the factors associated with the acceptance and rejection of remote care services. The findings provide insights into how digital healthcare may mitigate access barriers and reshape travel demand for medical purposes, particularly for underserved populations.
In the dining domain, the dissertation analyzes shifts in consumer dining preferences following the pandemic, with a focus on service channels (e.g., dine-in, pick-up, delivery) and restaurant segments (e.g., quick-service, coffeehouses, casual, and full-service). We identify distinct customer profiles for each channel, assess the influence of workplace location on dining behavior, evaluate patterns of substitution and complementarity across options, and analyze how compromise and variety-seeking drive decisions across restaurant categories.
Across all domains, this dissertation emphasizes the importance of understanding how technology, personal attributes, and the built environment collectively shape activity participation and travel demand. By bridging transportation engineering and behavioral science, the findings offer practical guidance to policymakers, planners, and industry stakeholders seeking to build equitable, efficient, and sustainable urban systems in a digitally enabled world.Civil, Architectural, and Environmental Engineerin
Evaluating Open Source LLMs for Claim Detection in Fact-Checking
No full textThe spread of misinformation makes fact checking tools more important than ever. ClaimBuster helps journalists and researchers by identifying factual claims that are worth verifying. To improve its accuracy, we trained state of art open source models: Microsoft’s Phi series and Meta’s LLama series.
We created ClaimHack, an extended dataset of check-worthy claims, to support fine-tuning and benchmarking. Both models were tested under multiple configurations and fine-tuned settings on TACC’s Lonestar6 system. Fine-tuned Phi-3 achieved the highest macro F1 score of 0.913, outperforming the DistilBERT (0.882), while LLaMA-3 reached a score of 0.899. Zero-shot versions demonstrated lower accuracy, emphasizing the importance of targeted fine-tuning for domain-specific tasks.
While fine-tuned Phi-3 and LLaMA-3 achieved higher accuracy, their increased inference time (0.06s to 0.09s/claim) made them less suitable for ClaimBuster’s public API, which journalists and researchers actively use. As a result, we retained DistilBERT in production to preserve real-time responsiveness (0.0023s/claim) while maintaining strong predictive performance.
This work highlights how combining advanced language models with TACC’s high-performance computing enables the development of impactful AI tools that support public trust, journalism, and information integrity.Texas Advanced Computing Center (TACC
U.S. Shale Oil & Gas - From Independence to Dominance Eagle Ford Shale: A Major Unconventional Oil & Gas Play in the U.S.
Get PDFBureau of Economic Geolog
Investigating rRNA Operon Diversity And Selection In The Long-Term Evolution Experiment
Get PDFBacteria typically harbor multiple ribosomal RNA (rRNA) operon copies in their genomes,
which are generally assumed to be functionally equivalent and subject to concerted evolution through gene conversion. However, recent work has demonstrated that sequence polymorphisms among Escherichia coli's seven rRNA operons produce substantial variation in translation capacity. Here, I use whole-genome sequencing data from the Long Term Evolution Experiment (LTEE) to investigate rRNA operon evolution across twelve E. coli populations over 50,000 generations. I developed a bioinformatics pipeline to track polymorphisms across operons and compared the observed temporal polymorphism dynamics to a stochastic neutral model of gene conversion. All twelve LTEE populations retained significantly more heterogeneity than predicted by neutral evolution alone. Additionally, only 19 new base substitutions arose and persisted, with none reaching fixation. These results suggest that there is not one optimal rRNA operon sequence. Rather, balancing selection maintains sequence heterogeneity among rRNA
operons under the LTEE conditions.Molecular Bioscience
Production Impact of Horizontal Well Orientation to Shmin Direction in the Permian Basin
Get PDFBureau of Economic Geolog