ScholarWorks@UMassAmherst
We are now able to accept submissions directly in ScholarWorks. For submissions that are not doctoral dissertations or masters theses, please log in with your NetID, click the + (plus) in to the top left corner, and select the Submit Research option.
Graduate students filing for February 2025 degrees: We are now accepting submissions directly to ScholarWorks. Directions for submissions can be found in this guide. Please email scholarworks@library.umass.edu if you have any questions.
Request forms appear to be functional. If you do not receive a reply to a submitted request, please email scholarworks@library.umass.edu.
This site is still under construction, please see our ScholarWorks guide for updates.
Recent Submissions
Publication Scalable Silicon Photodetector Arrays for Biomedical and Machine Vision Applications(2024-09)The development of advanced, scalable silicon (Si) photodetector arrays addresses critical challenges in both biomedical imaging and machine vision. Traditional fluorescence imaging methods in biomedical applications struggle with miniaturization, which is essential for ultimately implantable neural imaging and for point-of-care (POC) settings. Concurrently, machine vision systems face inefficiencies due to the physical separation of sensing and computing units, leading to high power consumption, data transfer issues, and escalating data storage challenges. In response to these challenges, this dissertation explores innovative solutions in two primary areas: fluorescence detection for biomedical applications and advanced in-sensor computing for machine vision applications. Firstly, in the realm of cell imaging, we focused on developing a spectrally filtered passive Si photodiode (PD) array for on-chip fluorescence imaging of intracellular calcium (Ca2+) dynamics. This device integrates a high-extinction-ratio spectral filter that effectively filters out strong excitation light, making the detection of weak fluorescence emission light possible. It captures both static and dynamic Ca2+ changes in C2C12 cells, demonstrating significant potential for pharmaceutical screening, cellular network studies, and potentially implantable neural interfaces. Additionally, we introduced an on-chip ratiometric aptasensing device to monitor cytokine dynamics. Utilizing pairs of sites-selectively spectrally filtered Si PDs functionalized with DNA aptamer probes, this device enables rapid detection of fluorescence changes due to aptamer-cytokine binding events at high sensitivity and specificity. Furthermore, the consistent and reliable performance across multiple runs makes it highly suitable for long-term POC diagnostics and therapeutic screening. Transitioning to the field of machine vision, the dissertation advances the concept of in-sensor visual processing by developing dual-gate amorphous-silicon photodiode arrays. These arrays are reconfigurable and can be gated to output either positive or negative photocurrent with zero power operation, thus enabling in-sensor computing in analog domain. By mimicking human retinal pathways, these bio-inspired arrays handle both static and dynamic visual information, allowing for multiplexed event sensing with sub-millisecond precision and edge detection of multiple objects. This innovative approach eliminates the need for extensive circuitry and the physical separation of sensing and computing units, significantly reducing power consumption and data transfer bottlenecks. The integration of these optoelectronics across different applications not only demonstrates their versatility but also underscores their potential to significantly impact areas ranging from pharmaceutical screenings to the development of autonomous machines. Each chapter builds upon the last, illustrating a clear trajectory towards achieving large-scale, efficient, and integrated systems that can adapt to both biomedical and intelligent environments. The potential for these advanced devices to revolutionize field-specific hardware and contribute to the advancement of intelligent systems is immense, paving the way for future innovations in these fields.Publication Regime Contestation and Racial Enfranchisement in the United States: from the Progressive Era to the Voting Rights Act of 1965(2024-09)Why do elites willingly bestow power upon excluded groups, thereby potentially reshaping the balance of power within society and political spheres? The democratization literature has grappled with these weighty questions since its inception as an organized field of study. While various explanations have been crafted to elucidate the dynamics surrounding working-class enfranchisement and women’s suffrage, our existing theories fall short of comprehensively addressing suffrage extensions to racialized minorities. Because racial enfranchisement often involves the extension of voting rights from a majority to a minority of the population, neither its determinants nor its consequences can be understood with frameworks developed to understand class-based extensions of the franchise from a minority to a majority (true for both manhood suffrage and to an extent viii also women’s suffrage). This dissertation seeks not only to carve out an analytical space for racial enfranchisement as both a distinct act of democratization and a novel concept with potentially profound implications for the basic concepts, focus, and assumptions of the democratization literature but also to study the case of the United States at two critical historical junctures: The Progressive Era and the complicity of major Republican players in the disenfranchisement of Black people in the Solid South, a case of failed enfranchisement, and the period between the 1930s and 1960s marked by the transformation of the Democratic Party from a staunched advocate of racial exclusion and suppression to the primary champion of a race reform, resulting with the enfranchisement of Black people through the Voting Rights Act of 1965. The former was marked by the entrenchment of a racially authoritarian regime despite the conducive electoral, structural, and political circumstances to the contrary whereas the latter saw an expansion of racial enfranchisement despite the absence of any clear electoral return and the presence of acute electoral risks for the Democrats as well as the existence of significant coalitions favoring racial exclusion. The dissertation demonstrates that racial enfranchisement takes place in a country under two circumstances: the emergence of an expansive vision of democratic peoplehood and its sway on actors in key governing institutions.Publication 2 Factors Challenging Faculty's Sense of Inclusion(2024-08-08)Pandemic-related caregiving burdens and health concerns have played a particularly large role, write Shuyin Liu, Dessie Clark, Laurel Smith-Doerr and Joya Misra.Publication COVID’s Lasting Impacts on Faculty Inclusion(2024-08-01)Think the pandemic is well behind us? Survey data shows feelings of inclusion have continued dropping as a result of it, write Laurel Smith-Doerr, Joya Misra, Shuyin Liu and Dessie Clark.Publication Exploiting Structures in Interactive Decision Making(2024-09)In this thesis we study several problems in interactive decision making. Interactive decision making plays an important role in many applications such as online advertisement and autonomous driving. Two classical problems are multi-armed bandits and reinforcement learning. Here and more broadly, the central challenge is the \emph{exploration-exploitation} tradeoff, whereby the agent must decide whether to explore uncertain actions that could potentially bring high reward or to stick to the known good actions. Resolving this challenge is particularly difficult in settings with large or continuous state and action spaces. For reinforcement learning, function approximation is a prevalent structure to manage large state and action spaces. However, misspecification of the function classes can have a detrimental effect on the statistical outcomes. These structured settings are the focus of this thesis. First we study the combinatorial pure exploration problem in the multi-arm bandit framework. In this problem, we are given $K$ distributions and a collection of subsets $\Vcal \subset 2^{[K]}$ of these distributions, and we would like to find the subset $v \in \Vcal$ that has largest mean, while collecting, in a sequential fashion, as few samples from the distributions as possible. We develop new algorithms with strong statistical and computational guarantees by leveraging precise concentration-of-measure arguments and a reduction to linear programming. Second we study reinforcement learning in continuous state and action spaces endowed with a metric. We provide a refined analysis of a variant of the algorithm of Sinclair, Banerjee, and Yu (2019) and show that its regret scales with the \emph{zooming dimension} of the instance. Our results are the first provably adaptive guarantees for reinforcement learning in metric spaces. Finally, we study a more fundamental problem of \emph{distribution shift}, where training and deployment conditions for a machine learning model differ. We study the effect of distribution shift in the presence of model misspecification, specifically focusing on $L_{\infty}$-misspecified regression and \emph{adversarial covariate shift}, where the regression target remains fixed while the covariate distribution changes arbitrarily. We develop a new algorithm---inspired by robust optimization techniques—that avoids misspecification amplification while still obtaining optimal statistical rates. As applications, we use this regression procedure to obtain new guarantees in offline and online reinforcement learning with misspecification and establish new separations between previously studied structural conditions and notions of coverage.
Communities in ScholarWorks
Select a community to browse its collections.