50107 research outputs found
Sort by
Decision Making in Graphical Models and Game-Theoretic Statistics
No full textThe goal of this thesis is to develop novel statistical and computational techniques for inference in graphical models and game-theoretic statistics, with a particular emphasis on tackling challenges in scenarios where data is expensive and not easily available. Recent breakthroughs in computational resources have dramatically reshaped the field of statistics, enabling the analysis of problems at scales that were once infeasible. The rise of high-performance computing clusters and the advent of specialized hardware like Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs) have redefined the limits of computational feasibility. This evolution has paved the way for tackling complex tasks, such as maximum a posteriori (MAP) inference, which were previously considered computationally prohibitive. In light of these advancements, the techniques developed in this thesis harness the power of modern computational resources while specifically addressing the challenges posed by costly and limited data collection.
This thesis advances the field in three major ways. First, we introduce a novel algorithm designed for MAP inference in general Bayesian factor models. Our approach employs Benders’ decomposition to iteratively incorporate constraints into the fully relaxed dual problem, systematically refining the solution space. This method not only provides a certificate of convergence but also preserves essential domain constraints, thereby delivering more accurate and reliable inferential outcomes.
Second, we extend the testing by betting framework to multi-agent environments in which multiple agents, each with their own betting strategy, operate under the umbrella of a single firm—the Principal—whose objective is to maximize overall wealth. Our extension unfolds in two distinct phases. Initially, we allow the agents to act independently while the Principal employs a dynamic wealth redistribution method that adjusts allocations at each time step. This proactive strategy minimizes exposure to significant fluctuations and helps prevent ruin, fostering consistent growth. We also enhance the model by accounting for the correlation among the agents’ betting behaviors. Leveraging an optimization approach inspired by modern portfolio theory, we enable the Principal to balance risk and return more effectively, resulting in a smoother and more stable wealth trajectory over time. Collectively, these advancements capture the nuanced interplay between individual incentives and collective dynamics, significantly broadening the framework’s applicability to complex, real-world scenarios where strategic interactions and effective risk management are critical.
Third, we tackle pivotal challenges at the intersection of sequential multiple hypothesis testing, high-dimensional variable selection, and genomic data simulation. Recognizing the critical need for efficient resource management in sequential testing, we introduce a Cost-Aware Expected -Wealth Reward (CAERO) framework. This novel approach is engineered to optimize sample allocation by incorporating finite-horizon constraints, thereby balancing immediate testing outcomes with long-term experimental goals in scenarios where data collection is costly. Complementing this framework, we present a novel method that fuses the model-X knockoffs technique with deep neural networks to accurately identify perturbation-responsive genes in large-scale biological experiments. This hybrid strategy enhances our ability to discern significant genetic responses in high-dimensional settings while rigorously controlling for false discoveries. Finally, we introduce SCSIM, a simulation tool designed to generate hierarchically structured single-cell and bulk DNA sequencing data. SCSIM addresses a critical gap by providing realistic synthetic datasets that are essential for developing and validating genomic analysis tools. Collectively, these contributions advance statistical methodologies and support robust, scalable analysis in modern genomic research.Doctor of Philosophy (Ph.D.)2026-05-1
PROFILING AND OPTIMIZING SOFTWARE PERFORMANCE AND MEMORY USAGE
No full textSoftware performance is crucial for both CPU-based programs and modern large language model (LLM) applications. Several factors influence performance, with memory usage being a significant one. However, memory-related performance issues are often associated with various dependencies, including low-level software components and hardware elements.
It is challenging to profile the impacts of those various dependencies precisely and efficiently. There are general-purpose profilers that utilize sampling methods to profile applications with low performance overhead. However, these tools mainly focus on the application itself and cannot identify issues arising from other dependencies.
This dissertation systematically analyzes performance bottlenecks and introduces effective optimization techniques. We introduce CachePerf, a cache miss profiler to identify cache misses; and MemPerf, a profiler to detect issues from the memory allocator. These profilers successfully identify most issues related to cache misses and memory allocation, achieving performance speedups of up to 3788% while imposing minimal performance overhead.
To analyze and optimize various types of software, we developed MemTrace, a memory analysis tool to profile dynamic memory management in autonomous driving (AD) software. We further propose Plasma, a framework for optimizing LLM inference with data transfer acceleration.
Overall, this dissertation proposes methodologies that improve profiling efficiency and allow for scalable, adaptive performance optimizations in various computing environments.Doctor of Philosophy (Ph.D.)2026-05-1
The second quarter century of UMass Linguistics: A brief overview and a web page archive
No full textLong-Term Impacts of Emergency Remote Teaching on K-12 Media Arts Educator Technology Use in the United States
No full textBetween March 2020 and May 2023, the COVID-19 pandemic triggered a national emergency and a rapid shift to Emergency Remote Teaching (ERT), requiring K–12 educators to integrate technology into their instruction in unprecedented ways. While extensive research has examined the immediate effects of ERT, there is less understanding of its long-term impact on educators' technology use, especially within specialized disciplines such as media arts. This dissertation examines how media arts educators in the United States engaged with technology before, during, and after ERT, using the PICRAT framework to analyze instructional practices. Through qualitative interviews, this study identifies patterns in media arts educators’ technology integration, barriers they encountered, and their post-ERT technology adoption. Findings suggest that—unlike trends reported in previous research—media arts educators consistently demonstrated high-level PICRAT integration before, during, and after ERT. While passive and replacement practices persisted, students remained actively engaged in creative, collaborative learning through educators’ use of amplified and transformational strategies. During ERT and continuing into in-person learning, the technology-related barriers participants faced shifted from resource-based challenges to those involving communication, engagement, and social-emotional support. By providing insights into the sustained effects of ERT, this study contributes to ongoing discussions on technology integration, professional development, and educational resilience, offering guidance for future emergency preparedness in K–12 education.Doctor of Philosophy (Ph.D.
Revisiting Branched Glycerol Monoalkyl Glycerol Tetraether Paleothermometry in Tropical Lacustrine Sediment Archives: Insight from Lake Malawi
No full textBranched glycerol monoalkyl glycerol tetraethers (brGMGTs) are a class of archaeal membrane-spanning lipids produced in the water column and anoxic sediment of lacustrine and marine settings. Like related glycerol dialkyl glycerol tetraethers (GDGTs), these compounds have been investigated for their potential application to paleoclimate reconstructions, particularly in tropical lakes. The mechanisms governing their production and structural distribution, however, have remained challenging to disentangle. Here we analyze brGMGTs from a section of the Lake Malawi Drill Core (MAL05-1) spanning several glacial-interglacial cycles to evaluate the relationship between these compounds and their changing environmental context. We find that many major brGMGTs exhibit notable sensitivity to changes in temperature and lake depth at Malawi, and we apply this insight to capture variability of brGMGTs produced predominantly in the anoxic water column through a new index (termed the ALX), which exhibits a moderately strong correlation with paleotemperatures and lake level at Malawi. We additionally revisit brGMGTs analyzed from modern East African lake sediments and find that the ALX strongly correlates with mean annual air temperature at these sites. These findings corroborate foundational insights from previous studies investigating brGMGTs as potential paleothermometers, while addressing some of the complications known to compromise previously-developed brGMGT temperature indices.Master of Science (M.S.)2028-09-0
Contemporary contexts of the self: architectural and geographic representations of roadblocks to self-construction in Marie Redonnet’s triptych of novels
No full textThis thesis examines the spatial contexts occupied by the narrators of Marie Redonnet’s triptych of novels (Splendid Hôtel, Forever Valley, Rose Mélie Rose) as a means for investigating those narrators’ halted constructions of self. Although structured in the fashion of traditional Bildungsromane, these novels withhold character development, the genre’s expected payoff. I argue that the geographic and architectural contexts of the novels contain representations of existential conditions that prohibit self-construction and are characteristic of the contemporary moment.
I use the parameters set out by Redonnet in her essay The Story of the Triptych as a barometer for the narrators’ self-constructions, the first and most pertinent of which has to do with coming to terms with one’s inheritance. My thesis, I argue, is in a broad sense a discussion of the narrators’ inheritances on literal and symbolic levels.
In an initial chapter dedicated to geographic context, I examine the dual character of the novels’ natural environments, which are by turns hostile and nurturing, as well as the symbolic nature of the novels’ landforms, which represent the alienation of the modern self geographically. Also in the first chapter is a consideration of how Redonnet’s natural symbolism operates on a metatextual level, speaking to her own path to self-discovery as an author who wished to renegotiate the novelist’s relationship to poetry.
In a second chapter, I consider the symbolism in representations of human-built architecture in the triptych. In a discussion on the ruinous quality of specific structures throughout the triptych, I analyze Redonnet’s portrayal of the degraded social context in which her narrators, understandably, struggle to gain subjectivity. I find that the dilapidation of certain buildings corresponds to the dilapidation of their analogous social institutions in the novels. Together, the ruinous states of religion, education, government and patriotism, and romantic and familial relationships point to an atmosphere of ruin, where both building, in the sense of creating new structures, and Bildung, in the sense of personal formation, are antithetical.
Overall, I find that Redonnet’s geographic and architectural symbolism foregrounds the challenging existential conditions of her narrators, painting a bleak portrait of the modern day.Master of Arts (M.A.
From Forum to Capitol Hill: Republicatn Decline in Rome and the Structural Tensions of 21st-Century American Politics
No full textWith the new presidential election over, the United States just entered a new era with dramatic political polarization, increasing social tension, and the growth of public anxiety. 2000 years ago, these familiar phenomena happened in another great civilization which experienced an autocratic, military control and that caused the republicans to transition to the Empire. This is Rome—-one of the most powerful civilizations in human history. Today, we may see another turning point in the United States. So, how do two countries lead themselves to face the similar turning point of the cross road of history? And since the loss of the Roman Republic had lots of similarities and differences with America, how will it warn the challenge of American democracy
Empowering Girls to Have a Choice: Investigating How Social Cognitive Career Theory Can Inform Middle School Computer Science Curriculums
No full textAs jobs in the computer and information technology occupations are projected to increase 11.5% between 2019 and 2029, it is imperative that we have a strong pool of candidates that is wholly representative of our population. The percentage of women in computer science occupations continues to be substantially less than for men, representing almost 1 in every 4 (24% as of 2023) employed in the field; it is therefore crucial to identify solutions that could inspire more girls to pursue careers in computer science. Increasing the number of women in computer science is not only equitable and a just action to take that fosters socioeconomic mobility but also ensures the U.S.A.’s economic competitiveness through diversifying perspectives and increasing creativity and innovation.
Middle school has been associated with critical developmental milestones that guide identity formation and potential career choices. Using a descriptive mixed methods approach, this study investigates how and if middle school CS curricula foster participation and engagement of girls in computer science. Social Cognitive Career Theory (SCCT), which is based on a triadic relationship between self-efficacy, goals, and outcome expectations, provides the foundation for identifying what components of CS curricula promote self-efficacy, career goals, and outcome expectations toward CS.Doctor of Philosophy (Ph.D.
Sleep Benefits Procedural Motor Learning in Older Adults With Additional Practice
Full text linkAging is associated with declines in both memory performance and sleep quality, raising questions about their potential connections. While younger adults consistently show sleep-dependent memory consolidation (SDC) benefits for procedural motor learning, findings in older adults have been inconsistent. The present study investigated whether additional training could enhance SDC in older adults, thereby mitigating age-related deficits in procedural motor memory consolidation. Young adults (18-24 years) and older adults (62-75 years) were assigned to standard training or overtraining conditions before completing an explicit serial reaction time task before and after a 12-hour interval of either wakefulness or sleep. Polysomnography was used to assess sleep architecture, including spindles and slow oscillation-spindle coupling, key neural mechanisms implicated in SDC. Behavioral results demonstrated significant SDC in young adults but not in standard-trained older adults, replicating prior research. However, overtrained older adults exhibited a significant sleep-related benefit in performance, supporting the hypothesis that additional training enhances encoding strength and facilitates consolidation. Additional analysis indicated that, once older adults’ encoding strength was sufficient to elicit sleep-dependent consolidation, weaker memories may be preferentially enhanced by sleep. Despite these behavioral improvements, sleep physiology did not differ between standard-trained and overtrained older adults, and sleep microstructure analyses revealed largely inconclusive associations between overnight performance changes and spindle activity or slow oscillation-spindle phase-amplitude coupling. Contrary to expectations, young adults showed weaker coupling during NREM2, and stronger coupling did not predict greater SDC in either age group. These findings suggest that encoding strength plays a critical role in determining sleep’s impact on procedural memory consolidation. Future research should further examine the interaction between pre-sleep encoding and sleep physiology to better understand the mechanisms underlying SDC for procedural memory in aging populations.This work was funded by NIH R01AG040133 (PI: Spencer).Master of Science (M.S.