Electronic Theses and Dissertations - Open Access

Permanent URI for this collectionhttps://hdl.handle.net/20.500.12588/2237

This collection contains electronic UTSA theses and dissertations (ETDs), primarily from 2005 to present. The collection is not comprehensive; search the UTSA Library Catalog for a complete list of UTSA theses and dissertations.

All of the ETDs in this collection are available to the general public. However, authors are able to request an embargo. Embargoed ETDs will not be downloadable until after their embargo expires.

Authors of these ETDs have retained their copyright while granting UTSA Libraries the non-exclusive right to reproduce and distribute their works.

Former students are invited to broaden access to their thesis or dissertation by making it available in the Open Access collection. To initiate this process, or if you have any questions about the ETD collection, please contact rrpress@utsa.edu.

Browse

Now showing 1 - 20 of 479

Reconnecting Our Fragments through Ceremony: An Indigenous Archaeology Comparative Ceramic Analysis in the Coca-Nahua Pueblo of Mezcala, Jalisco, Mexico.
(2024) Figueroa Alcantar, Jesus
This masterâ€™s thesis revitalizes archaeological inquiry from an Indigenous perspective in the Lake Chapala Basin of West Mexico through a comparative ceramic analysis of registered and non-registered ollas (jars) at the Mexkalâ€™lan Community Museum and olla rim sherds from the Island of Mezcala, Jalisco Mexico with the Ph.D. dissertation of Margaret Nell Bond published in 1971. Due to historical exclusion of Indigenous perspectives in archaeology, I use my lived experience in Indigenous ceremony as a guiding theoretical framework in our process of interaction and interpretation of the material culture of our ancestors from Lake Chapala. I apply methodology from Indigenous archaeology that incorporates interviews with Indigenous Coca-Nahua elders in the town of Mezcala, Jalisco, Mexico in order to center local Indigenous Knowledge in the process of knowledge production. Therefore, this thesis is an initial conversation with the community of Mezcala on previous archaeological reports in the region, an explanation of the process of this ceramic analysis, with the research goals of understanding the regional context of these ollas in order to expand our understanding of dating and the chronology of our different tribal cultures of the Lake Chapala Basin. The results of this comparative ceramic analysis brings the out-of-context ollas of Mezcala into closer context with other medium-sized ollas of similar form and style from the IxtÃ©pete-El Grillo phase of the Valley of Atemajac (present-day Guadalajara, Jalisco) and the Lagos phase of Los Altos de Jalisco, both dating to the Classic period (300-900 CE). Recognizing the challenges to access to western forms of higher education and financial resources, I will donate two refurbished laptops with this thesis translated to Spanish and all cited references to Cesar Hesiquio â€œPapayoâ€ Santiago de la Cruz and Noel Contreras Garcia from Mezcala, to encourage them and the community to continue to investigate and write their own history, and our collaboration as Indigenous colleagues and scholars.
Biogeochemical Controls on Arsenic Mobility within Hyporheic Zone Sediments
(2024) Varner, Thomas Scott
Arsenic contaminated drinking water is a global concern, specifically in the Bengal basin where millions rely on arsenic contaminated groundwater for drinking purposes. Currently, the primary process believed to be responsible for the high dissolved arsenic is the microbially-mediated reductive dissolution of arsenic-bearing iron-oxides. Recent studies suggest that the interactions between oxygen-rich surface water and iron-rich groundwater in the hyporheic zone precipitates abundant iron-oxide minerals which sequester arsenic. The objective of this dissertation is to investigate the comprehensive role of hyporheic zone processes on the cycling of arsenic in sediment along the Meghna River, Bangladesh, and the Hooghly and Beas Rivers in India. The inorganic and organic chemical properties of the riverbank sediments were evaluated and the resulting biogeochemical processes influencing arsenic mobility within the hyporheic zone were determined. The findings show three distinct hyporheic zone scenarios which impact the fate of arsenic through differing biogeochemical processes. Along the Meghna River, a shallow silt layer, rich in labile organic matter, promotes arsenic mobility in the riverbank by fueling the microbially-mediated reductive dissolution of arsenic-bearing iron-oxides. Along the Hooghly River, surficial sands and minimal organic matter permits the precipitation of arsenic attenuating iron-oxides, maintaining low arsenic concentrations in the riverbank. Along the high-energy Beas River, a low residence time and oxic conditions prevents the microbial reduction of oxides, allowing for efficient transportation of As-bearing minerals to the underlying deltas. Together, this research provides a comprehensive analysis on the solid-phase properties of hyporheic zone sediments influencing the fate and transport of arsenic.
Ceramic Production and Distribution: Testing a Middle Preclassic Ceramic Economy in the Mopan River Valley, Belize
(2024) Villarreal, Alessandra Noelle
This project develops a testable model to study Middle Preclassic (1000 - 300 BCE) ceramics from two sites in the Mopan River valley. Early Xunantunich and San Lorenzo may typically be considered opposite ends of the sociopolitical scale. Early Xunantunich boasted monumental construction as early as 800 BCE and likely served as a location of communally integrative feasting and ritual performances. Visible from this early monumental site was the Middle Preclassic community of San Lorenzo, represented archaeologically by few known perishable dwellings and a large, buried chultun. Through a lens of ritual economy, I initially hypothesized that Early Xunantunich was a node of intercommunity gathering activities that hosted community members from surrounding settlement sites, including San Lorenzo. Community participants likely brought offerings in ceramic vessels, which were ultimately left at the site after the event. This was tested through a compounding, multistage methodology that assessed the modelâ€™s material correlates using type-variety, petrography, and Neutron Activation Analysis (NAA), and ethnoarchaeological collaboration. This methodology confirmed that Early Xunantunich was a node of an inter-community, and indeed interregional, network that likely gathered at the site for rituals, feasts, and the construction of Structure E-2-3rd. However, it also revealed that residents of San Lorenzo likely also performed complex intra-community rituals at home, supplied by community-specific potters. These findings shine new light on the sociopolitical and economic relationships of the Mopan River valley in the Middle Preclassic period, and warrant more exploration into our assumptions about the interactions between monumental sites and surrounding â€œhinterlandâ€ communities.
Revealing "La Gran Ocultadora": Frida Kahlo at the Intersection of Art and Linguistic Repertoire as Representative of the Cultivation of Identity
(2024) Diaz, Kandice Nikkole
The present dissertation investigates the linguistic features of Mexican artist Frida Kahloâ€™s register that serve as acts of identity meant to index various parts of her identity. I explore how Kahlo constructs and performs her identity through her writings as a result of language contact, in relation to mexicanidad, and as a means to mark her as a member of the speech community of Mexico City. The methodology used in this study is a combination of corpus stylistics and close reading. Data collection began with a manual scan of the corpus and utilized a digital tool (Internet Archive) to perform KWIC searches to locate occurrences of specific linguistic features. Once the linguistic features were located, close reading was used to ascertain Kahloâ€™s rhetorical purpose and the contextual meaning of each piece of data. The corpus used for examination purposes is comprised of Kahloâ€™s writings, to include her letters and diary, all of which were analyzed in Kahloâ€™s mother-tongue to ensure reliability of conclusions made based on the analysis. I argue that Kahloâ€™s language use presents an invaluable resource in which to examine her construction and performance of identity in addition to the visual aspects of this performance. I also argue that Kahloâ€™s writings display her alignment with ingroups and specific ideologies as deliberate rhetorical choices meant to garner a desired response from her audience, while other aspects of her register occur as an example of natural language use. Thus, this study adds to the limited scholarship on identity work based on a written corpus, fills the gap on linguistic studies conducted on Kahlo beyond a biographical focus, and offers an expanded application of Le Page and Tabouret-Keller's â€œacts of identityâ€ theory.
Practicing Care in the Borderlands: Family, Fulfillment, and Everyday Experience among Older Migrant Caregivers in the Rio Grande Valley
(2024) Ramirez, Jessica Jazmin
This thesis will analyze the everyday lives of older Mexican women who have migrated from Mexico to the Rio Grande Valley (RGV) and who have experiences with family caregiving to better understand how their cultural ideologies of care, gender, and kin relations shape their care practices. This research responds to the call from anthropology and migration studies to integrate the perspectives and experiences of women who have migrated to correct the implicit assumption that men are always central figures (e.g., Brijnath 2009; de la Luz Ibarra 2002; Ibarra 2022; Raijman and Schammah-Gesser 2003; Yarris 2017b). I explore these older Mexican womenâ€™s gendered migration experiences to gain insight into how they establish their own place-making within their homes in the RGV. Additionally, this research seeks to understand how these older Mexican women experience everyday caregiving within the home in light of their gendered social roles and obligations shaped within and by kin relations (Mohanty 1998). Within this discussion I was able to explore these older Mexican womenâ€™s local cultural ideologies of gender, care, and kin relations and how they influence their care practices and everyday realities in the RGV. Through this, I also examine the sense of fulfillment and stressful obligation that accompany family caregiving among the older women in my project. Broadly, I argue that highlighting these cultural narratives from older Mexican women in the RGV contributes to a deeper understanding of the diverse and complex local cultural ideologies of gender, kinship, and care practices within the context of life histories of migration.
Resource-aware Tiny Machine Learning for Battery-less System
(2024) Islam, Sahidul
Powerful machine learning algorithms have been increasingly designed to achieve better accuracy, which however require a great amount of data and computing power relying on centralized cloud services. This generates a series of problems such as high cost, privacy issue, carbon emission, and low quality of service due to large data transmission overhead. Recently, a counter-trend of tiny machine learning that shrinks those large models for IoT devices is promising in addressing those issues. There are billions of IoT devices around the world that serve for various common daily-life applications such as human activity recognition, voice command recognition, face recognition etc. It's necessary to implement tiny machine learning algorithms on such IoT devices to enable those applications. However, there are fundamental challenges in designing tiny machine learning models due to the resource constrains such as the limited computing capability, memory, and energy of IoT devices which are mostly based on microcontrollers. Typical microcontrollers have low computing power (e.g., 1-16 MHz) and are equipped with small memory (e.g., hundreds of KBs). Besides, battery-powered devices naturally have a limited standby time. Although battery-less IoT devices that harvests energy from ambient environment can work sustainably, the power provided by the energy harvester is low and has an intrinsic drawback of instability since it varies with the ambient environment. Computations on such devices are interrupted frequently and become intermittent. To address these challenges, we propose several software/hardware co-design methodologies to efficiently implement tiny machine learning on battery-less devices. First, we introduce an end-to-end framework to accelerate machine learning model while achieving efficient intermittent computation. The whole framework is divided into three module such as resource-aware DNN pruning, accelerator enabled embedded software, intermittent inference module. We demonstrate how these modules interact to achieve improved performance. Second, to adapt to the varying environment and harvesting power, we propose environmental adaptive machine learning models for low-power energy harvesting battery-less devices. A co-exploration framework is proposed to search multiple machine learning models with shared weights. We further propose on-device implementation architecture to efficiently execute such shared-weight models. A run-time model extraction algorithm is proposed that retrieves individual model from the shared source. Third, to achieve multi-tasking machine learning in varying environments, we present a scalable multi-tasking machine learning framework that generates a single unified machine learning model which exhibits the flexibility to adapt with the varying environment while performing multiple machine learning tasks.
Structured Segment Rescaling with Gaussian Processes for Parameter Efficient Convnets
(2024) Siddiqui, Bilal
We study methods to transform exiting Neural Networks (NNs) into more parameter efficient variants using a novel mechanism for structured pruning. The pruning method Structured Segment Rescaling (SSR) functions as downsampler on model dimensions and utilizes rescaling modifiers. These modifiers act on a segment, which is a logical group of identically dimensioned blocks. We study the behavior of SSR in Convolutional Neural Networks (ConvNets) and its generalization from heuristics into a well-defined framework. Our novel structured pruning starts at model instantiation where we begin with heuristics that explore radical segment rescales. The rescales construct ConvNets with varied segments that can have new dimensions where some or most blocks and channels may be pruned away. In contrast to iterative unstructured pruning, SSR is significantly more aggressive, requires a single train cycle, and purposefully targets parameter banks to completely avoid tensor sparsity. Since sparse ten- sors require similar compute to dense tensors, our circumvention of sparsity uniquely places SSR amongst the prior work. SSR significantly reduces computation as measured in General Matrix Multiplications (GeMMs); we show that optimized SSR modifiers can achieve up to a 5X lower GeMM compute load. The modifiers that rescale model segments in SSR are also augmented with an optimization step using a low cost Gaussian Process (GP). The GP serves to approximate the optimal modifiers using an initial set of depth and width modifiers that enumerate only extreme rescales. ConvNets constructed from these initial modifiers are named the sentinels. The sentinels coarsely explore the modifier space and provide the data bedrock for training GPs. We utilize the CIFAR datasets and ResNets to validate our findings. SSR only requires 10¹ GPU hours to yield efficient new ConvNets that can facilitate edge inference. Over 10⁵ ConvNets may be derived from any typical ConvNet and these ConvNets need only be trained if their GP predicted accuracies show that they are viable candidates for power limited devices. Our sentinel models drop parameter count by over 65% and improve latency by 3X. We then further optimize for better modifiers with GP modeling, and show that up to 80% structured parameter reduction is possible. We observe that both depth and width modifiers can significantly reduce parameters. We also note that only depth modification decreases latency because fewer blocks means less serial computation. Lastly, applying depth and width modifiers simultaneously to segments significantly increases ConvNet compression. We demonstrate that <1% accuracy degradation and >90% parameter reduction is possible when modifiers are jointly optimized.
Continuing Nixon's Method: Republican Engagement with Hispanic Voters
(2024) Mendoza, Ian Robert
The research presented this thesis examines Richard Nixonâ€™s method for increasing his Hispanic vote share and investigates how subsequent Republican presidents sought to emulate his approach. To analyze the extent to which future Republican presidents followed Nixonâ€™s method, a descriptive analysis will be conducted on its implementation during the initial campaigns, first terms, and re-election campaigns of Ronald Reagan, George H.W. Bush, George W. Bush, and Donald J. Trump.
A Case Study of Leadership Practices in a Dual Language Program
(2024) Almanza, Olivia
The purpose of this qualitative study was to investigate the perception of the principal and teachers regarding what transformational leadership practices meet the needs of EBs in a DLI PK-8th grade campus in South Texas. The research site was selected for having 40% emergent bilingual students. The campus principal has been implementing a comprehensive 80:20 dual language immersion program for more than five years, and the campus has received an â€œAâ€ rating from the Texas Education Agency based on State of Texas Assessment of Academic Readiness (STAAR) assessments more than once. Data were collected from the research site by utilizing Seidman (2006) interview series which consist of three rounds of structured interviews from the principal, two interviews from two bilingual teachers, and one from the focus group which consisted of three bilingual teachers. Following SaldaÃ±aâ€™s (2021) coding process, the data were collected, categorized, and given a code aligned to the transformational leadership theory. The main themes that emerged on the leadership practices used to implement a comprehensive dual language program in a PK-8th grade campus are the following: (1) Build a Shared Vision through Stakeholder Involvement; (2) Model School Practices and Advocate; (3) Build relationships, Empower, and Distribute Leadership; (4) Facilitate Teacher Reflection for Student and School Improvement. The findings indicate that the principal set structures to transform a traditional school to a dual language immersion campus. He engaged stakeholders in building the vision of the school and modeled practices that he expected his staff to emulate. One of the core leadership practices that he implemented was collaboration and used it to engage staff in working as a team as they met and discussed school progress. He built positive relationships with staff, students, and families. He inspired his subordinates to explore new approaches, empowered them, and increased capacity. He also created space for students to meet with him so they could express campus improvements that needed to be made based on their lens. Transformational leadership practices have had a positive effect in a dual language campus in obtaining high levels of performance of teachers which ensured high academic achievement of students.
Robustness and Dependability of Deep Learning Models for Real-World Applications
(2024) Wang, Zhiwei
Robustness and dependability of Deep Learning (DL) models are critical for real-world DL-based applications. In this dissertation, we explore real-world threats affecting three DL-based applications: a smartphone-powered, computer-vision-based system for Power Wheelchair Intelligent Assistive Driving (PWC IA-Driving), a gene expression-based Deep Neural Network (DNN) for cancer-type prediction, and an efficient and intelligent attack detection in Software Defined IoT Networks. Subsequently, we propose methodologies to enhance the robustness and dependability of these applications. Firstly, we develop the PWC IA-Driving system, aiming to enable the safe, hands-free operation of a Power Wheelchair (PWC) with reduced user attention required in indoor environments. This system alleviates the burden on disabled individuals and reduces their stress. Our goal is to offer an affordable and practical solution that can be seamlessly integrated into existing PWCs. The system utilizes a customized and pre-trained ResNet-based model on a smartphone to interpret driving commands from real-time imagery captured by the phone's camera. These instructions are then transmitted to the PWC via a control interface connected to the smartphone. We have developed a prototype of this assistive driving system on a Pixel-6 Android phone and tested its feasibility on a mobile robot as a proof-of-concept. To ensure the mobile robot can navigate safely at reasonable speeds with minimal user intervention, we employ various techniques to enhance robustness and dependability. These include model explanation to increase confidence, data augmentation to improve accuracy for unseen scenarios, model distillation and quantization to enhance robustness against possible adversarial attacks, and utilization of on-device acceleration devices to improve response time and tolerance to low-confidence predictions, among others. Secondly, we delve into the threats posed by adversarial attacks on Deep Neural Networks (DNNs), where adversaries can manipulate DNN outputs by crafting small, carefully designed perturbations to the inputs. These attacks present significant challenges to the practical deployment of DNNs. In this dissertation, we investigate a variety of defense methods against adversarial attacks on gene expression-based DNNs for cancer-type prediction. We propose a novel method called "segment patching" to mitigate the effects of adversarial perturbations. Segment patching effectively replaces perturbed input data segments with clean segments from the training dataset based on Euclidean distance. Our experiments demonstrate that this method maintains model prediction accuracy against adversarial attacks, particularly strong attacks. More importantly, the segment patching method poses a significant challenge to adversaries attempting to generate adversarial examples. Additionally, we explore the application of Fast Fourier Transform to transform input data into the frequency domain before feeding it into the DNN model. This approach aims to further obscure model gradients, making gradient-based attacks more difficult. Our findings suggest promising strategies for enhancing DNN robustness against adversarial attacks. Thirdly, we explore the dependability issues surrounding Deep-learning based abnormality detection for IoT coupled with Software Defined Network (SDN) technologies. With the increasing integration of IoT devices into various domains such as smart buildings and critical infrastructure protection, their limited capabilities pose significant security vulnerabilities, especially when coupled with SDN technologies to provide flexible services. In this study, we concentrate on Random Forest (RF) machine learning models and scrutinize the impact of different feature sets (e.g., IPs and ports) on the detection accuracy for various attacks. Our aim is to enhance dependability through two main approaches: firstly, evaluating the effects of RF configurations (specifically forest size and tree depth) on detection accuracy and runtime overheads to improve response time by reducing forest size; secondly, generating a substantial amount of trusted labeled data by simulating attacks within our SDN environment. Our findings indicate that RF demonstrates high detection accuracy with the selected feature sets across different attacks. Furthermore, even with reduced forest sizes (e.g., fewer trees or shallower depth), the detection accuracy of RF only experiences a slight decrease, allowing for significant reductions in runtime overheads and thereby enhancing response time.
Do You Really See Me: Understanding and Enhancing Educatorsâ€™ Recognition and Response to Middle School Studentsâ€™ Social and Emotional Needs
(2024) Gonzales, Roseann
This qualitative multi-case study engaged the testimonios of three educators who openly shared their social and emotional experiences from their middle school years. The central inquiry, â€œWhat valuable insights can you offer from your middle school experiences to help other educators recognize and respond to the social and emotional needs of their students?â€ directed their reflection on their journeys. The primary focus was to investigate how these experiences can offer help into recognizing and addressing the social and emotional concerns of other educatorâ€™s students, explicitly utilizing the Ethics of Care theoretical framework as the basis for their evolution. This framework centers on building nurturing connections with teachers and staff and is a blueprint for fostering encouraging connections with students. Demonstrating care through modeling can foster mutual respect among teachers, staff, and students, as Noddings (2005) suggested. Through data collection, open coding, theme coding, overall analysis, and interpretation, the results created a conceptual framework termed "The Chain of Social and Emotional Trauma." This framework offered another lens to visualize the pathway of emotional understanding, enabling educators to identify and address potential struggles faced by students. The insights from testimonios shared in two off campus sessions sparked candid reflections on the educators' experiences. Common themes, including vulnerability, trauma, silent suffering, coping mechanisms, resiliency, and social and emotional resources, emerged with profound sincerity. These themes served as inspiration for constructing the conceptual framework, offering a transition into the potential struggles and challenges middle school students may be facing today. This study was meant to increase awareness among educational leaders, empowering them to strategically plan future initiatives to meet students' needs through advocacy and implementing school-based mentorship programs.
Optimal Retrieval for Continual Learning at Scale
(2024) Hickok, Truman
In recent years, deep neural networks have emerged as extremely scalable machine learning models. These models are often trained on billions of samples; however, as gaps in models' skill sets are revealed, naively continuing training after the expensive initial training phase leads to rapid forgetting of past tasks and reduced transfer to new tasks. Continual learning research is concerned with developing methods to counteract these effects, allowing models to continue training over data streams of indefinite length without overwriting existing representations. One of the most widely used approaches in continual learning is referred to as replay. Replay methods support interleaved learning by storing past experiences in a replay buffer. Although there are methods for selectively constructing the buffer and reprocessing its contents, there is limited exploration of the problem of selectively retrieving samples from the buffer. Current solutions have been tested in limited settings and, more importantly, in isolation. Existing work has also not explored the impact of duplicate replays on performance. In this thesis, we propose a framework for evaluating selective retrieval strategies, categorized by simple, independent class- and sample-selective primitives. We evaluated several combinations of existing strategies for selective retrieval and present their performances. Furthermore, we propose a set of strategies to prevent duplicate replays and explore whether new samples with low loss values can be learned without replay. In an effort to match our problem setting to a realistic continual learning pipeline, we restrict our experiments to a setting involving a large, pre-trained, open vocabulary object detection model, which is fully fine-tuned on a sequence of 15 datasets.
LABOR MARKET INTERGRATION OF AFRICAN IMMIGRANTS: THE US IN A COMPARATIVE PERSPECTIVE
(2024) Okabe, Lydia Mabel
This dissertation provides a thorough understanding of the labor market integration specifically the employment and occupation outcomes of African immigrants in comparison to other immigrant groups and native-born individuals in both the United States and key European countries, specifically Spain and Italy. The study examines factors shaping their integration, such as employment outcomes, educational selectivity, and occupational preferences. Adopting a comparative perspective, the research seeks to unveil both the shared characteristics and distinctive features in the integration experiences of African immigrants compared to other immigrant groups in the United States and Europe. This study employs two datasets, namely IPUMS ACS (2016-2020) and IPUMS International (2000, 2001 & 2011), to investigate the employment patterns of African immigrants relative to other immigrant groups and native-born individuals. Additionally, it assesses the employment outcomes of African immigrants in the United States, Italy, and Spain, aiming to discern the influence of host country factors and educational selectivity on employment trends. Furthermore, the study examines the impact of education and host country on the occupational engagement of African immigrants in high-skilled occupations within Italy, Spain, and the USA. Gender-specific analyses are conducted to account for potential disparities in migration patterns and labor market participation between males and females. Through logistic regression analyses and the inclusion of interaction effects involving education, place of birth, and host country, the findings reveal noteworthy variations in the employment dynamics of African immigrants compared to other immigrant cohorts and native-born individuals in the United States. Additionally, surprising insights emerge regarding the employment patterns of African immigrants in Italy, Spain, and the USA, along with their involvement in skilled labor across the three examined countries.
Leading While Oppressed: Black Women Leading Urban High Schools in Texas
(2024) Dixon, Sharene Lynn
This study undertook a thorough investigation into the complex experiences of Black female principals leading urban high schools in Texas, with a primary objective of spotlighting the pervasive underrepresentation of Black women in high school principalship. Through rigorous examination, the research evaluated the multifaceted challenges faced by secondary Black female principals as they navigate to and through the position. Beyond merely highlighting adversities, the study illuminated the remarkable triumphs, setbacks, and daily realities of being a Black female principal amidst educational landscapes with inequities and biases. In addition to uncovering hurdles, the research captured the diverse experiences of Black female principals, exploring their unique strategies, perspectives, and insights. By immersing itself in their narratives, the study provided a comprehensive understanding of their leadership journey, offering valuable insights into how they navigated the complex intersectionality of race, gender, and educational leadership. Furthermore, it shed light on the resilience, innovation, and unwavering determination exhibited by these lady leaders as they strove to sustain success in pivotal roles shaping the academic, social, and emotional well-being of their students and communities. Qualitative Sista Circle Methodologically is a method used to employ a blend of research, surveys, and Sista Circles rooted in Black Feminist Thought and intersectionality. Through these avenues, the research uncovered emergent themes such as the journey through principalship, challenging stereotypes, navigation within high school leadership, advocacy, and Social Justice Leadership. Additionally, subthemes like Imposter Syndrome, Unorthodox Elevation, Uplifting Others, and Overcoming Trauma provided nuanced insights into the experiences of Black female principals in educational leadership.
Adventures in Catalysis: Development of Transition Metal-Catalyzed Reactions for C-X Bond Formations and High-Resolution Mass Spectrometry Tools for Probing Catalytic Reaction Mechanisms
(2024) Silva Villatoro, Roberto Napoleon
The dissertation contained herein will chronical the development of new Pd- and Ni-catalyzed C-X bond forming reactions and the initial development of a new direct-injection high-resolution mass spectrometry (DI-HRMS) tool for detecting organometallic complexes in cross-coupling reactions at synthetically relevant concentrations. Chapter 2 focuses on the development of a Pd-catalyzed, chemoselective O-selective benzylation of 2-quinolinones and 2-pyridones. These classes of N,O-heterocycles are deemed ambident nucleophiles, i.e. they have 2 nucleophilic sites on the same molecule. Though N-alkylation is well established, selective O-alkylation has been largely dependent on the use of stoichiometric Ag salts. In collaboration with the Chemical Process Development team at Bristol Myers Squibb (BMS), we have developed a novel Pd-catalyzed system capable of selectively alkylating the oxygen atom of 2-quinolinone using benzyl bromides as the electrophilic partner. Mechanistic experiments identified a surprising Xantphos mono-oxide Pd(II) Î·1-benzyl complex as the resting state of our catalyst and responsible for such high chemoselectivity. Expansion of this chemistry to 2-pyridone, a related heterocycle, has proven a much larger task than initially thought. Through optimization, new biphasic reaction conditions have been identified as a promising alternative to achieve a similar chemoselectivity as was observed with 2-quinolinones. Further mechanistic experiments have begun to probe the C-O bond forming step. Chapter 3 stemmed from a curiosity from the work in Chapter 1. The Xantphos mono-oxide Pd(II) Î·1-benzyl complex was first identified through DI-HRMS. Without this piece of data, the discovery of this intermediate would have been greatly hindered. Given the capabilities of high-resolution mass spectrometers to detect compounds even at low concentrations, we set out to explore other canonical Pd-catalyzed cross-coupling reactions in a similar manner. So far we have explored both Sonogashira reactions and C-O coupling reactions. We have successfully detected multiple on-cycle complexes as well as some surprising off-cycle species. Current works are focused on exploring several protocols for Heck reactions and Buchwald-Hartwig aminations so explore the effects of different conditions on the detectable species. Chapter 4 follows a second collaboration with the Chemical Process Development team at Bristol Myers Squibb. Though Ni-catalyzed C-N coupling has been studied for decades, their applicability to a broad range of substrate classes and large scale reactions has left much to be desired. Using the power of HTE at BMS, we set out to establish a Ni catalyst to unify the reactivity and substrate scopes of previous incarnations while making dramatic improvements to functional group tolerability and scalability. Herein, we have demonstrated a Ni-catalyzed system capable of the C-N coupling of aryl chlorides with anilines and aliphatic amines under markedly milder conditions to those previously reported. We have demonstrated the unique reactivity of our Ni-catalyzed system and are currently pursuing expanding the electrophile scope.
Replay for Online Continual Learning in Spiking Neural Networks
(2024) Patel, Raghav
Continual learning, also called lifelong learning, is the process of accommodating newly learned information with prior information. A major challenge neural networks face in continual learning, is remembering what was learned previously. This issue is referred to as catastrophic forgetting. When the same neural network is trained on a new batch of data, it will perform poorly when tested on previously learned data. This problem is far from solved, and it becomes apparent that tackling this issue is a difficult balancing act: the neural network must be plastic enough to learn, but at the same time stable enough such that it does not forget what it learned. This investigation explores different replay approaches within spiking neural networks. We show that SNNs benefit in a similar manner as do traditional ANNs when simple replay is used and show that adding regularization can improve accuracy when using small replay buffer sizes.
Apply Reinforcement Learning to Control Traffic Signals by Action Space Optimizations
(2024) Figueroa, Mauricio Eli
Vehicle traffic systems can be improved by optimized traffic signal control. Performance metrics including accumulated waiting time and total fuel consumption are affected by different configurations of traffic signal controllers. This thesis describes a workflow for defining effective traffic signal controllers utilizing reinforcement learning. Configuration of components for reinforcement learning are explained, with a focus on phase set and action space definitions, to result in performance metrics improvement. A combinatorial search method where traffic signals were configured with different viable phase sets is presented, to identify individual phases of interest or concern. An improved phase set, now action space for the reinforcement learning agent(s), was then constructed composed of top performers from the combinatorial search. Analysis results show that reinforcement learning agents benefit from having an action space defined from highly ranked phases and learn to produce combined positive behaviors from different phase sets. At peak hours, morning, noon, and evening, of vehicle traffic simulation, optimized phase sets and action spaces produced a reduction in accumulated waiting time and fuel consumption.
Runtime Verification of Real Time Properties and Bounded Model Checking in the Thorium Reactive Programming Language
(2024) Baldor, Kevin Scott
Writing software that reacts to external events in real time is notoriously difficult. Some of that difficulty comes from the languages and frameworks that we use to write that software and I argue that a declarative language can mitigate some of that challenge by ensuring that the full definition of each reactive element exists in only one place in the code. To support real time specifications, we introduce a novel monitoring algorithm for several useful subsets of Metric Temporal Logic (MTL) and describe the time and space complexity of each. We then embed the monitor for the past-time subset -- due to its linear space and time complexity as well as the guarantee of timely results -- into a declarative reactive framework so that a programmer can define real time properties and respond if they are violated. Finally, we present a declarative language, textit{thorium}, for defining reactive software. It employs many of the operators of Functional Reactive Programming (FRP) but with a novel mechanism for reconfiguration that trades expressiveness for improved readability. We present its semantics, describe their encoding into a satisfiability modulo theories solver, and evaluate the performance of its bounded model checking -- First on a toy re-configurable processing pipeline and then a more realistic application. We conclude with a discussion of opportunities for improvement of the model-checking performance and discussion of the interaction of the semantics of functional reactive programming and metric temporal logic.
Very Large Community College Last-Dollar Promise Program: Impacts on School-to-College Enrollments
(2024) Orona, John
This dissertation examines the impacts of a last-dollar promise program on high school-to-college enrollments at a very large community college (VLCC) district in Texas. The research intends to add to the current literature on the growing popularity of promise programs at two-year community colleges. Specifically, this quantitative study examines first-time-in-college (FTIC) students at high schools with low school-to-college enrollments. It examines the impacts of a promised scholarship opportunity offered by the VLCC at no cost or low cost. The research questions aim to examine if a VLCCâ€™s last-dollar promise scholarship in a metropolitan area increases high school-to-college enrollments. In addition, the demographics of the student sample are reviewed to explore the enrollment effects of the last-dollar scholarship by the observable student groups of gender, prior college credit attainment, and ethnicity. Difference-in-differences (DD) and event-study (ES) methods were used to analyze the effects of the VLCC promise program on high school-to-college enrollments. DD analysis is a widely used methodology to review policy or program effects by sampling a population before and after implementation (Schwerdt & Woessmann, 2020). Event studies complement the DD methodology results by providing further insight into the treatment effect over time (Li & Katri, 2023). The findings in this study are similar to other studies that examined promise programs at community colleges as explained in Chapter 3. This investigation showed that the promise scholarship offered by the VLCC increased high school-to-college enrollments at each promise-eligible high school by 26 to 28 students, resulting in a 4% to 5% increase in enrollments at the promise-eligible high school for students attending the VLCC. In perspective, the overall impact showed an increase in the total enrollment rate at the VLCC from all promise-eligible high schools, ranging from 15% to 25% (log). The results suggest that the VLCC had a statistically significant impact on high school-to-college enrollments. The demographics showed significant increases in enrollments for both genders, with females increasing by 15 more students, while males increased by 13 more students enrolled at the VLCC from promise-eligible high schools compared to schools that did not have the promise-scholarship opportunity. There was an overall 14% (log) increase for females, while males showed a 10% (log) increase in students at the VLCC from all promise-eligible high schools. For FTIC students with no prior earned college credits, the results showed an increase of 27 more student enrollment at from eligible high school, representing a 14% (log) increase at the VLCC from all promise-eligible high schools. In contrast, students with prior college credits failed to show changes in enrollment at the VLCC. African American and White student enrollment saw a statistically significant increase of five and eight more student enrollments at schools with promise scholarships, respectively, while Hispanic students showed 14 more students. This translated to African American students showing an increase in overall enrollments at the VLCC by 21% (log), White students by 23%(log), and Hispanics by 9% (log) from promise-eligible high schools. Groups classified as Asian, and Other failed to show any increases in enrollments at promise-eligible high schools.
Leveraging Latent Fields for Accurately Attributing Model Behavior
(2024) Payne, Ethan Robert
Large 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.

Browse

Recent Submissions