33760 research outputs found
Sort by
Queering the Timeline: Imagining Queer Futures from Weimar Germany to 21st Century Switzerland
No full textThe queer communities in 1920s Weimar Germany and in the 2020s in countries such as the United States, Germany, and Switzerland share similar conditions for imagining queer futures existing outside the realm of heteronormativity. Both periods show signs of a fluctuating relationship between, on the one hand, perceived progress, increased visibility, and acceptance of the queer community, and, on the other, political and social backlash adhering to heteronormative constraints and attempting to foreclose the queer futures lying ahead. In this thesis, I examine how queer communities a century apart imagine queer futures. Through close readings of two illustrated magazines aimed at the German-speaking queer community, Die Freundschaft (1919-33) and Die Freundin (1924-33), and the 2022 novel Blutbuch by Swiss author Kim de l’Horizon, I reveal how and which queer futures are proposed by texts that lend a voice to queer sexual and gender identities. In addition to the thematization of queer sexual and gender identities, a common characteristic in all three texts is the use of assemblage to express queer futurities. The depictions of queer futures within these texts reject notions of heteronormative temporality, while challengingreaders to imagine a queer utopia that flouts restrictive, exclusionary, and harmful norms and policies upholding heteronormativity in society. I show how queer futures proposed in both magazines and in the novel utilize desire, political visions, and the reorganization of temporalities, spaces, and forms to encompass logics that transcend the principles structuring the heteronormative timeline. Thus, the visions for queer futures proposed in these texts, published a century apart, reveal the importance of aesthetic and literary depictions of temporalities that challenge readers to imagine a queer utopia, one in which marginalized groups such as the queer community no longer face the risk of futures or existences being extinguished
Machine Intelligence-Accelerated Discovery and Design of Sustainable Functional Materials
No full textDeveloping advanced sustainable materials – such as biodegradable plastic alternatives, high-performance biobased packaging, and multifunctional aerogels – is critical, yet remains hindered by slow trial-and-error methods and one-variable-at-a-time experimentation. Traditional approaches struggle to navigate large, multidimensional design spaces and meet complex, multi-objective performance targets. This dissertation introduces a data-driven, closed-loop framework that integrates collaborative robotics, machine learning prediction, and simulation tools to accelerate the discovery and optimization of sustainable materials with tailored functionality.
The robotics/ML-integrated workflow is demonstrated across three distinct material systems. First, a machine intelligence–accelerated approach—combining collaborative robotics and ML prediction—enabled the discovery of all-natural plastic substitutes using four GRAS (Generally Recognized As Safe) components. The workflow, integrating active learning and predictive modeling, supported two-way design tasks by enabling both forward prediction and inverse design of biodegradable and biocompatible materials that match the optical, thermal, and mechanical properties of conventional plastics. Molecular dynamics simulations further revealed key strengthening mechanisms within the optimized formulations.
Second, the robotics/ML-integrated workflow is scaled to a broader formulation space of 23 natural/GRAS components to develop sustainable biobased packaging aimed at improving postharvest preservation. By integrating robotic automation, active learning, ML predictions, Density Functional Theory simulations, and life cycle assessment, the workflow enables the discovery of a library of biobased nanocomposites with enhanced mechanical resilience, tunable transparency, antimicrobial functionality, and environmentally benign compositions.
Third, the robotics/ML-integrated workflow incorporates generative modeling to further accelerate the development of mixed-dimensional aerogels with customizable structural and mechanical properties. Robotic automation of aerogel fabrication, mechanical testing, and SEM imaging is used to generate a high-quality dataset. Predictive models estimate key mechanical and microstructural features, while a diffusion-based generative model synthesizes realistic SEM-like images. These synthetic structures informed finite element simulations, enabling multi-physics performance evaluation and multi-objective property optimization.
Collectively, this work establishes a scalable and unconventional ML-driven materials development platform that significantly shortens the discovery-to-optimization cycle, paving the way for rapid innovation in sustainable functional materials
Faith Communities As A Solution To The Affordable Housing Crisis
No full textFinal report for PLCY400: Senior Capstone (Spring 2025). University of Maryland, College ParkThis paper examines how faith institutions can use the Facilitating Affordable Inclusive Transformational Housing Zoning Text Amendment (FAITH ZTA) to build affordable housing in Montgomery County, Maryland. We examine past literature on the means and methods other communities have undergone to build affordable housing on church-owned land across the countries. We also conducted interviews with faith institution leaders in Montgomery County. Our results suggest that faith leaders are interested in what the ZTA provides and believe that their churches’ missions closely align with the amendment's provisions. Given the data collected from our literature and our independent interviews, we propose three recommendations for how faith leaders should approach the FAITH ZTA: (1) Partner with nonprofit organizations and developers; (2) Join and/or learn from Enterprise Faith-Based Development Initiative; and (3) Engage in community outreach events to address pushback concerns.Montgomery County, M
Minimalist Sensing Toward Ubiquitous Perception
No full textThis thesis explores minimalist sensing, a design philosophy prioritizing simplicity and efficiency in sensor technology to capture essential information for various applications, including robotics, environmental monitoring, and wearable technology. By focusing on streamlined functionalities, these sensors avoid the complexities and costs of more elaborate systems, offering practical solutions under resource constraints. My research emphasizes developing low-power, miniaturized systems that integrate seamlessly into both urban and natural environments, enhancing ubiquitous perception without the encumbrance of complex technologies.
I explore three main areas: low-power and miniaturized acoustic direction-of-arrival (DoA) estimation, ultra-low-power spatial sensing for miniature robots, and a single frequency-based tracking interface for voice assistants. The contributions include a novel low-power DoA estimation system using 3D-printed metamaterials, an innovative spatial sensing system for mobile robots using a single speaker-microphone pair, and a comprehensive voice and motion tracking interface that operates on a single frequency. This work is aimed at establishing a pervasive perception network that offers continuous, reliable data while minimizing energy use and infrastructure demands, potentially revolutionizing real-time monitoring and responsiveness in diverse settings.
A critical aspect of achieving minimalist perception is the integration of machine intelligence and computation. By leveraging advanced algorithms and computational techniques, we can bridge the gap in minimalist perception, making it both feasible and efficient. Machine learning and signal processing algorithms enhance the accuracy and functionality of simplified sensor systems, allowing them to perform complex tasks without the need for sophisticated hardware. For instance, intelligent data processing enables low-power sensors to extract meaningful information from limited data inputs, reducing the need for extensive sensor networks. By incorporating these computational strategies, we can push the boundaries of minimalist sensing, enabling the creation of smart, resource-efficient perception systems that are capable of operating in diverse and challenging environments
Extending Theories of the Graph Matching Problem and Its Variants
No full textGraphs serve as powerful tools for modeling complex real-world relationships, making reliable statistical inference on graphs a crucial task. A fundamental problem in this domain is the graph matching problem, which seeks to align node labels across graphs while minimizing structural and feature discrepancies. In this thesis, I investigate algorithms for the graph matching problem and one of its key variants, the subgraph detection problem.
I develop theoretical frameworks that leverage signals from a clustered, vertex-aligned collection of graphs to accurately recover node labels in a newly shuffled network and classify this new network into one of the clusters. Additionally, I propose a novel approach for detecting multiple instances of a noisily embedded template graph within a large background graph. Furthermore, I explore the relationship between the anonymization time and the mixing time of a specific class of Markovian noise applied to graph edges. Beyond these contributions, I address several related challenges in graph matching and its related optimization algorithms, offering new insights into their theoretical and practical aspects.
To validate our methodologies, I provide rigorous theoretical justifications and conduct extensive experiments using both simulated and real-world network data. These findings demonstrate the effectiveness of the proposed approaches, bridging the gap between theoretical advancements and practical applications in graph inference
A MACHINE LEARNING APPROACH TO PREDICTING HIGH-RISK IRRITABILITY TRAJECTORIES ACROSS THE TRANSITION TO ADOLESCENCE
No full textAdolescence is a sensitive developmental period that presents a crucial opportunity for early intervention to mitigate risk for future psychopathology. Transdiagnostic symptoms are important indicators of risk. Irritability, characterized by proneness to anger, frustration, and temper outbursts, is a transdiagnostic symptom associated with negative mental health outcomes in youth and adults. Prior work has characterized youth with varying levels of irritability across different developmental periods and identified irritability trajectories that signify high risk. Data were from the Adolescent Brain Cognitive Development (ABCD) Study, which is a 10-year longitudinal study that tracks the brain development, cognitive skills, physical health, and psychosocial functioning of a large, national sample starting from preadolescence. The baseline sample consisted of 11,862 9-10-year-old preadolescent youth. Irritability was parent-rated at baseline, 1-year, 2-year, 3-year, and 4-year follow-ups on the Child Behavior Checklist (CBCL) irritability index. Latent class growth analysis (LCGA) was used to determine developmental trajectories of irritability. Four machine learning approaches were applied to develop predictive models for youth irritability trajectories. The baseline (preadolescent) variables covered a wide range of domains (youth psychopathology, youth physical health, neurocognitive abilities, psychosocial environment, demographics, parent psychopathology, and structural neurobiology). LCGA identified four distinct irritability trajectories: persistent low irritability (n = 8692, 73.28%), moderate irritability and decreasing (n = 1083, 9.13%), low to moderate irritability and increasing (n = 1449, 12.22%), and chronic high irritability (n = 638, 5.38%). The machine learning models demonstrated strong performance in detecting probability of being in the chronic high irritability trajectory, but poorer performance in classifying the two irritability trajectories that were characterized by change in irritability. In addition, the machine learning models showed strong performance in differentiating the persistent low irritability trajectory from the other trajectories. The top predictors indicated that the youth psychopathology domain produced the most relevant predictors. The machine learning models also highlighted several novel predictors for irritability trajectory which merit further research. Furthermore, behavioral and clinical variables outperformed the structural neurobiology variables in predicting irritability trajectories. The present study provides a foundation for future development of models to predict irritability trajectories during a critical developmental period for early intervention
EVIDENCE ON PUBLIC AND PRIVATE INVESTMENTS IN COMMUNITIES
No full textIn this dissertation, I study how communities respond to local policy and investment changes, using evidence from public libraries and place-based policies to understand how public and private interventions impact community behavior, institutional outcomes, and local economic development.
In the second chapter, I estimate the effect of eliminating overdue fines on library usage and finances. Since 2010, a growing number of public libraries have eliminated fines for overdue materials, citing concerns about access. Using panel data of Illinois public libraries from 2007 to 2019 and a staggered difference-in-differences framework, I find that removing overdue fines increases library visits and circulation, with suggestive evidence that it is driven by libraries in lower-income communities. I find no evidence of negative financial impacts, as lost fine revenue is small relative to total operating income. These findings suggest that fine elimination can improve efficiency in library use without compromising financial sustainability.
In the third chapter, I study the effects of philanthropic grants from the Bill and Melinda Gates Foundation on public libraries, examining their impact on computer spending and access, library service availability and use, and library spending. Between 1997 and 2003, the Foundation awarded grants to over 4,500 library systems across the US to support public access to computers. Using panel data and a staggered difference-in-difference design, I find limited evidence of any impact to libraries across outcomes considered. The grants may have increased the number of public access computers, but consistent pre-trends in outcomes make exact interpretation of this and all outcomes challenging. At best, the results suggest that philanthropic investments of this variety may address infrastructure deficits in the short run but have more limited effects on long-term public service usage.
In the final chapter, I study the impact of the New Markets Tax Credit (NMTC) program, a place-based policy, on local development outcomes. Established in 2000, the NMTC program provides tax credits to incentivize private investment in economically distressed communities. Using comprehensive data on NMTC-financed projects and a fuzzy regression discontinuity design, I find limited evidence that the program meaningfully changed local economic conditions.
While some outcomes show statistically significant changes in some years, these effects are not consistent over time or easily interpreted. Heterogeneity analyses very tentatively suggest that if there are effects, that they may be stronger in higher-vacancy and lower-income tracts, but the variability in estimates and the strength of the first stage warrant caution. Overall, the findings highlight the challenges of identifying local effects of place-based investments and suggest that NMTC may, in many cases, have limited or no measurable impact
Fairness and Uncertainty in Graph Matching
No full textGraph matching is a fundamental computational problem with a broad range of applications across different domains, including resource allocation, e-commerce, kidney exchange, ad placement, and ride-sharing. While the classical maximum matching problem is well-understood, many real-world applications impose additional practical considerations such as uncertainty in the graph structure and fairness requirements. In this dissertation, we consider variations of matching that incorporate these considerations, and describe some key results for these problems.
First, we examine online bipartite matching with stochastic edges, where the vertices of one side of the partition arrive sequentially and the algorithm knows only the \emph{probability} with which each of its incident edges exists. Upon the arrival of a vertex, the algorithm must \emph{probe} the edges one by one to determine existence, committing to match the first one found to exist. Arriving vertices may further have a \emph{patience} value (the maximum number of its incident edges that may be probed), which may itself be stochastic (in which case the algorithm has only distributional knowledge of the patience). We present a framework for these problems and derive results under adversarial, known IID, and ``prophet'' arrival models, so long as the single-arrival case can be solved. To this end, for the single-arrival case, we show an optimal algorithm for a ``constant hazard rate'' model, and a -approximation under general patience distributions. We also consider a two-sided arrival model and demonstrate a constant-factor approximation for this model. Finally, we present some negative results related to these settings.
We then consider \emph{offline} stochastic matching for bipartite graphs, and give a (1-1/\euler)-approximation for this problem when the vertices of one side all have patience 1. Our approach has connections to contention resolution schemes, which we then investigate further. We describe results for contention resolution schemes for the matching polytope under two different arrival settings: adversarial order and random order; and for both general graphs and bipartite graphs.
Finally, we study proportionally fair matching, where we are given an \emph{edge-colored} graph, and we must find a matching of maximum possible weight while ensuring that the proportion of edges in the matching of any given color is between some specified parameters and . Due to strong hardness results for the exact version of this problem, we consider a slightly relaxed probabilistic version instead. Here, we allow for small violations in the proportional fairness constraints, and achieve a constant factor approximation with probabilistic guarantees on the amount by which the proportionality constraint is violated
“SOMEBODY TO TALK TO”: HUMANIZING FRONLINE LABOR AT THE NATIONAL BUILDING MUSEUM
No full textThe museum labor movement has grown significantly in the last ten years, with museum workers increasingly agitating around issues of inadequate pay, discrimination, job precarity, unsafe working conditions, and more. Unionization and other organized responses to these injustices have dominated discourses about this movement. Frontline workers, or museum workers that deal directly with the public, hold the least structural power and are most vulnerable to exploitation, particularly given the dangers of the COVID-19 pandemic. However, narratives among museum workers and scholars often paint frontline workers as unhappy, burnt-out victims, or overlook them entirely. This dissertation ethnographically explores the experiences of frontline workers at the National Building Museum in Washington, DC and seeks a holistic, nuanced understanding of frontline workers in a museum without a union. In addition to examining challenges frontline workers face, it examines how workers create and derive joy from their work and each other, and how they come together in solidarity outside of organized labor agitation. Physical and social space are explored as key components in understanding frontline workers’ experiences. This dissertation argues that despite a challenging labor environment and a disconnection from the organized museum labor movement, frontline workers at the National Building Museum leveraged joy, solidarity, and space to create a workplace that worked for them. In doing so, this work resists labor narratives victimizing frontline workers, highlights how workers support each other informally, and uplifts frontline workers as multidimensional people enjoy and care deeply for each other and their work
ClarifAI - Detecting Breast Cancer with AI
No full textBreast cancer is one of the most commonly diagnosed cancers worldwide and a leading cause of death, especially among women. The lack of early detection, negatively impacts survival rates, as treatment is more effective in the early stages of the disease. This research project investigates the use of artificial intelligence, specifically machine learning, for the early detection of breast cancer using mammogram imaging. By training a network model on a dataset of labeled roughly 1,000 mammogram images, titled normal, abnormal, cancerous, and benign, the goal was to develop an AI model capable of distinguishing between cancerous and non-cancerous mammogram scans. Initial model accuracy was around 25%, indicating random prediction levels. However, after refining the model and adjusting the training process, accuracy increased to approximately 37%, demonstrating the model’s ability to learn and improve. Although this performance is not yet suitable for clinical use, the results confirm the model’s potential and offer a foundation for future improvements. Challenges such as overfitting, data bias, and limited data availability, were explored as their impacts were shown during the research process. The findings support the role of AI as a valuable tool in advancing medical diagnostics and improving patient outcomes through earlier breast cancer detection