80264 research outputs found
Sort by
Rethinking Maximum Likelihood Estimation
Get PDFThis thesis is a culmination of my doctoral work which addresses the issue of bias in statistical parameter estimation, with a particular focus on deep generative models. I propose a novel alternative to Maximum Likelihood Estimation, most popular method for estimating parameters, and show how my proposed method mitigates bias, reduces overfitting and the overrepresentation of high frequency events, increases the representation of low frequency data, is more stable in a MAD setting. This thesis details how this method can be solved analytically, used with existing estimators, and implemented in deep learning settings with hypernetworks. Inspired by this method, I propose a new grading scheme that incentives students to report their true beliefs, and show how the proposed method evaluates cognition and metacognition simultaneously. This thesis justifies both methods on the basis of the Bayesian interpretation of probability, and details how such an interpretation works as an extension of logic, defending this interpretation against rival views. Finally, this thesis addresses logic itself (and why Logic is important to Electrical and Computer Engineering), providing the necessary background from Aristotle's axioms to Linear Algebra, to statistical parameter estimation
Meantime Design: Architecture, Infrastructure, and Housing Struggles in Post-Apartheid Cape Town
No full textThree decades after the end of South Africa’s apartheid regime, Cape Town remains a deeply fractured city with little prospect for transformation or systemic change. It is still one of the least integrated and most segregated cities in the world. Based on long-term ethnographic research conducted in Cape Town between 2018 and 2024, this dissertation examines how architects, housing activists, and other built environment professionals have developed alternative designs and participatory initiatives to address the urgent needs of residents living in flood- and fire-prone informal settlements and occupations. Designing in and with the meantime was a common denominator in many of these architecture initiatives. While liberatory promises of the post-apartheid era have been repeatedly deferred, the “meantime” emerged as a temporal horizon for alternative designs, including public housing upgrades, adaptive reuse models, and incremental building typologies. By emphasizing small-scale, imaginative, and participatory models and upgrades that negotiate alternative urban futures and improved infrastructural conditions, this dissertation complements anthropological research on architecture, infrastructural temporalities, and housing struggles. Through an analysis of various “meantime designs” and their characteristics in each chapter, it also contributes to a deeper understanding of design interventions that may reflect potential urban futures yet perpetuate South Africa’s most exclusionary housing landscape
6.4 A Statement of Shared Stewardship (Video)
Get PDFThis entreaty was created as part of The Spirit of Asilomar and the Future of Biotechnology summit (February 23-26, 2025) in Pacific Grove, CA
Mechanical Properties of Two-dimensional Nano-composites and Oxides
No full textTwo-dimensional (2D) materials have attracted enormous interests owing to
their extraordinary properties due to their atomic-level thickness and robust in-
plane atomic bonding, positioning them as promising materials for advanced
technological applications across electronics, photonics, sensing, energy storage,
and structural composites. However, their practical applications have been hindered
by intrinsic brittleness, susceptibility to defects, and relatively low fracture
toughness. This thesis systematically investigates intrinsic and extrinsic toughening
mechanisms, along with anisotropic fracture properties, in selected novel 2D
materials and composites to enhance their mechanical robustness and reliability.
The intrinsic toughening mechanisms are explored through detailed studies
on monolayer amorphous carbon (MAC) nanocomposites, investigating how
structural heterogeneities (crystalline and amorphous domain) influence fracture
resistance. In-situ scanning electron microscopy (SEM) tensile testing with
molecular dynamics (MD) simulations provides comprehensive insights into
fracture processes and toughening behaviors.
Extrinsic toughening strategies are investigated through two-dimensional
covalent organic framework sandwich structures, demonstrating significant
improvements in fracture toughness. Additionally, anisotropic fracture behavior is
studied in monolayer titania nanosheets, emphasizing the role of crystallographic
orientation and defect distributions in determining mechanical performance.
Overall, this thesis provides fundamental insights into fracture mechanics
and toughening mechanisms in 2D materials, offering practical strategies for
designing mechanically robust and reliable nanocomposite systems. The findings
not only advance the fundamental understanding of 2D material behavior but also
expand their potential applications in next-generation engineering technologies
Expansion and Shockwave Development in Ultracold Neutral Plasmas with an Initial Exponentially Decaying Density Distribution
No full textUltracold neutral plasmas (UNPs), which are formed by photoionizing a cloud of cold atoms, have been a powerful platform for studying a wide variety of plasma phenomena. Two areas of interest include expansion into vacuum and the development of shockwaves. Previous UNP experiments have used plasmas with an initial Gaussian density distribution. UNPs with a radially exponentially decaying density distribution, or exponential UNPs, can be formed by photoionizing atoms out of a purely magnetic trap. This dissertation studies the expansion in free space of exponential UNPs as well as the development and characterization of shockwaves in exponential UNPs.
The free space expansion of an exponential UNP is compared to the well-known self-similar hydrodynamic expansion of a Gaussian plasma. Similar scaling laws describe the size evolution well for both cases. The evolution of the plasma size and velocity show a characteristic ion acoustic time scale, indicating that the plasma is generally well-described by a hydrodynamic description. While the exponential UNP is predominantly in the hydrodynamic regime, heating at the central peak for low-density, high-electron temperature suggests significant local non-neutrality at early times.
This thesis presents the first observation of a shockwave in a UNP. The evolution of the density and velocity for exponential UNPs show signs of wave steepening. A significant density and velocity jump over a narrow region develops. These occur at the same location in the plasma. Additionally, a large spike in the ion temperature occurs at this front. The relative ion velocity across the front modestly surpasses the local sound speed.
The initial conditions were varied to characterize the development of shockwaves. Varying the density and electron temperature did not impact shock formation for the ranges used. Changing the shape caused the shocks to disappear as the density gradient decreased. This suggests that the electron thermal pressure gradient plays a significant role in the shock formation. The establishment of shock formation in UNPs opens a new avenue of research for UNPs
What is Forgotten Can Change: Reckoning with Sites of Public Memory in Texas
No full textWhat is Forgotten can Change: Reckoning with Sites of Public Memory in Texas develops critical Texas studies as a method, field, and critical apparatus that combines feminist critical regionalism and public memory studies to interrogate how places are made, mythologized, and calcified. Building from and critiquing traditional Texas Studies—which attempts to create a coherent, singular, and universal Anglo-Texan identity positioned as official history—critical Texas studies challenges this coherence through a complication of its static, often jingoistic, understandings of place-making, memory, and region. I contend that Texas Studies has become normalized in such a way as to seldom be recognized as an academic or political project, instead being presumed as objective truth, enduring at public memory sites, in K-12 curricula, state legislation, and popular perceptions of Texas both within and beyond the state. Critical Texas studies disrupts this coherence by reflexively examining sites of public memory—such as the Alamo, the Texas-Mexico border, the Texas Ranger Hall of Fame and Museum, and College Park Memorial Cemetery—to reveal how what I term white possessive logics, which are rooted in masculinist entitlement and ownership, structure dominant narratives of place and belonging. These logics, while not inherently violent, encourage violence in order to uphold whiteness and maleness as foundational elements of Texas’ identity. Spanning from the Post-Emancipation period to the present day, my project critically examines key historical moments including the settlement of Coahuila y Tejas, the Texas Revolution, the Republic of Texas, the monument and memorial mania of the early twentieth century, the Civil Rights era, NAFTA, and post 9/11—demonstrating how racial politics have shape public memory. Ultimately, this work advocates for a communal, decolonial, and anti-racist approach to place-making that destabilizes dominant myths and opens space for justice and equity
A multiple artificial viscosities approach to entropy stable discontinuous Galerkin methods
No full textEntropy stable discontinuous Galerkin (DG) methods display improved robustness for problems with shocks, turbulence, and under-resolved features by enforcing an entropy inequality. Such methods have traditionally relied on entropy conservative (EC) fluxes that are computationally expensive to evaluate. An alternative approach for enforcing an entropy inequality is through a minimally dissipative artificial viscosity. We review how to construct an artificial viscosity formulation and extend this approach to artificial viscosities with multiple parameters (e.g., viscosity and thermal diffusivity). Through the method of Lagrange multipliers, we determine simple analytical expressions for optimal viscosity parameters. We compare this to the case of a single monolithic viscosity parameter for different 1D and 2D problems, and show that the proposed method allows users to more precisely target specific physical phenomena while retaining robustness for general problem settings
Innovative Statistical Methods for Improving Clinical Trial Design and Personalized Medicine
Get PDFClinical trials are pivotal for advancing medical treatments, yet they are often burdened by high costs, lengthy timelines, and substantial failure rates, particularly in oncology. These challenges underscore the need for innovative approaches to improve trial efficiency and facilitate personalized treatment strategies. This thesis introduces novel statistical methods and computational tools designed to address these issues by optimizing clinical trial design and enhancing personalized medicine.
Chapter 1 presents a multiple-dose randomized clinical trial design aimed at identifying the optimal dose that maximizes the benefit-risk tradeoff. We generalize the standard definitions of type I error and power to accommodate the unique aspects of dose optimization and derive a decision rule, along with an algorithm, to determine the optimal sample size. This design, referred to as MERIT (Multiple-dosE RandomIzed Trial design for dose optimization based on toxicity and efficacy), incorporates Bayesian interim rules to allow for early trial termination due to toxicity or futility. Simulation studies show that MERIT offers favorable operating characteristics, with a sample size of 20 to 40 per dosage arm generally providing reasonable power and type I errors to ensure patient safety and benefit. To support implementation, we provide the MERIT design software, available at www.trialdesign.org.
In Chapter 2, we introduce the self-adapting mixture (SAM) prior, a novel method for dynamically incorporating historical data into ongoing trials. SAM priors are data-driven and self-adjusting, adaptively weighing the informative or non-informative prior component based on evidence of prior-data conflict. This results in dynamic information borrowing, with SAM priors demonstrating desirable properties in both finite and large samples and achieving consistency in information borrowing. Additionally, SAM priors are computationally efficient, data-driven, and calibration-free, minimizing the risk of data dredging. Numerical studies show that SAM priors outperform existing methods in addressing prior-data conflicts effectively. We have developed an R package, “SAMprior”, and a web application, both freely available on CRAN.
Chapter 3 introduces a novel Bayesian adaptive design for avatar-driven cancer clinical trials, which uses mouse or laboratory animal grafts to create personalized tumor models to guide patient treatment. This design adaptively selects the optimal treatment for each patient by jointly modeling avatar tumor growth and patient clinical outcomes, including toxicity and progression-free survival, to capture the interaction between avatars and their source patients. A latent-class model is used to account for the possibility that avatars may only be effective surrogates for treatment effects in a subset of patients.
In Chapter 4, we explore intratumor heterogeneity (ITH) in cancer, a key factor in treatment response, using multi-region gene expression data. We propose ICeITH, a Bayesian hierarchical model that incorporates cell type profiles as prior knowledge to decompose mixed bulk data while accounting for within-subject correlations among tumor samples. ICeITH quantifies ITH by assessing the variability in targeted cellular compositions. We also demonstrate ICeITH's ability to stratify patients based on their ITH scores and link these estimates to survival outcomes in two non-small cell lung cancer datasets