80264 research outputs found
Sort by
Anticipating Obsolescence
No full textStudying the American Data Center through a series of details intended to be deconstruted for future use once the building's function has become technologically obsolete
1.5 Identifying and Addressing the Risk of the Environmental Release of Organisms — Engineered or Natural
No full textThis entreaty was created as part of The Spirit of Asilomar and the Future of Biotechnology summit (February 23-26, 2025) in Pacific Grove, CA.The environmental release of both engineered and non-engineered organisms for Biotechnologies Beyond Conventional Containment (BBCC) offers unique solutions to pressing global challenges, including the prevention of soil degradation, the attenuation of nitrogen pollution, the replacement of harmful pesticides and herbicides, the remediation of anthropogenic contaminants and ‘forever chemicals’ mitigation. An evaluation of impacts, both positive and negative, rather than arbitrary prohibitions, is crucial for advancing the responsible use of organisms intentionally released into the environment. The history of biological interventions demonstrates that organisms have successfully contributed to agriculture, pollution remediation, ecosystem restoration, waste upcycling, and pest control, yet their full potential remains constrained by regulatory hurdles that do not fully account for modern scientific advancements. At the same time, some releases serve as cautionary tales, having caused harm due to a lack of regulation and monitoring. Unlike chemicals released to the environment, organisms — particularly those designed or selected for specific functions — can be managed with built-in safeguards, ranging from physical and genetic containment strategies to controlled ecological interactions to mitigate risks while maximizing benefits. Advancements in precision engineering, computational modeling, and real-time monitoring technologies now allow for unprecedented accuracy in tracking, assessing, and controlling the environmental impact of released organisms — capabilities inaccessible when recombinant DNA technology first emerged 50 years ago. Many regulatory structures were developed decades before today’s explosion of biological knowledge and insight was even imaginable. This resulted in our current policies that have become restrictive, limiting the deployment of innovative and promising biological solutions. A new approach to risk analysis is now needed that accounts for changes in science, and in society, which assesses the environmental release of natural, evolved, and engineered organisms based on their functions rather than their origin or how they were developed. By modernizing these frameworks to emphasize continuous assessment, real-world data collection, and adaptive risk assessment and management, stakeholders can create a regulatory pathway for the sustainable, responsible, and evidence-based integration of environmental biological technologies
Symbolic Approaches for Boolean Synthesis
No full textBoolean synthesis is the problem defined as the procedure to construct solutions for unknown variables in a given specification in Boolean formula as a conjunction of constraints describing the relationship over known and unknown variables. Formally, the problem consists of two parts, the identification of full, partial, and nullary realizability for the input domain, and the construction of the functions for the unknown variables. As a fundamental problem with applications in circuit design, formal verification, and temporal logic synthesis, in the last few decades there have been algorithmic solutions in relevant works using mainly search-based and also symbolic approaches, especially those using Binary Decision Diagrams (BDDs), an acyclic directed graph that maps the solutions for a boolean formula to its paths. Scalability challenges also rose as exponential memory blowups occurred in handling large-scale problems using BDDs, but it also becomes one of the most interested measurement that values the potential to solve larger problems for industrial purposes.
Zero-suppressed Decision Diagrams (ZDDs) offer more compact representations for sparse and large-size conjunctive normal form (CNF) formulas, which decomposes input formulas into factored components that are represented by tree decompositions, with improved performance in realizability checking and witness function synthesis construction. Dynamic programming(DP) framework, which also shows capability and potential in model counting works, improves the operations such as existential quantifications and conjunctions by utilizing graded project-join trees. This dissertation introduces approaches that take ZDDs and dynamic programming to address these limitations and add solvers to the portfolio of promising industrial synthesis tools. Three algorithms are proposed as symbolic approaches for boolean synthesis, with their corresponding tools ZSynth, DPSynth, and DPZynth. The first approach is monolithic ZDD-based, shows better performance in compilation and realizability checking, but has a tie dependent on benchmark families with BDD-based tool. The second approach taking BDDs and DP shows a general better performance over non-DP tool and machine learning-based tool even if taking DP overhead into consideration. The third approach has better general picture in time performance, overcomes the planning overhead, and has a scalability potential on some benchmarks. By empirical evaluations, we show the existence of unique features for the proposed approaches, and evaluate their strengths as necessary additions to the portfolio of industrial solvers. The crux of the portfolio does not rely on a particular always best solver, but is proved to be in need of future works that adjusts the selection based on characteristics of a problem
Historical Analysis of Lifetime Justice Involvement of Harris County Youth (Executive Summary)
No full textBetween 2010 and 2019, more than 42,000 youths became involved with the Harris County juvenile justice system. However, by the time they “aged out” at 17, their experiences with the juvenile justice system showed striking differences. This report provides a detailed representation of the multiple ways in which youths interact with the Harris County Juvenile Justice System. It examines the overall level of involvement of youths with the system, as measured by the number of contacts or referrals accumulated throughout their lives. It also investigates how those contacts, and the ways in which the system reacted to them, changed as some youths became repeatedly involved with the system. Furthermore, it explores the extent to which information available at the time of a youth's first contact with the system may or may not help identify youths who could benefit from preventative and rehabilitative programs. Our analysis takes a historical view of the Harris County Juvenile Justice System. Instead of focusing on the youths in the system right now, we use data for youths who were involved at some point in their adolescence but have already “aged out” of the system. Specifically, we analyze the histories of all youths who were born between 2000 and 2002 and had their first contact when they were between 12 and 16 years old.2 Thus, the analysis in this report reflects the system as it was experienced by youths who are no longer under its jurisdiction. This implies that any recent changes to the juvenile justice system will not be captured by this report, or will only be captured to the extent they were experienced by some of the youths who aged out very recently. The historical data demonstrate that most justice-involved youths have only one contact with the system, while a small number of youths account for a disproportionately high share of referrals. At the same time, consequences become increasingly severe as these same youths become repeatedly involved with the system. To the extent that youths of color, particularly Black youths, are more likely to be detained and to receive relatively more severe dispositions than white youths during their initial contacts, these patterns have a disproportionate effect on them. Overall, these findings point to a need for targeted, early interventions and further, rigorous research to understand how we can better identify youths at risk of entering this cycle. Such interventions could potentially contribute to the reduction of racial disparities in the way the system treats and affects different groups of youths
Generative AI in the Age of Synthetic Data: Challenges and Solutions
No full textThe increasing scale of generative models in artificial intelligence (AI) has led to a growing reliance on synthetic data, raising fundamental challenges for model stability and data quality. Training new generative models on synthetic data produced by prior generations can create a self-consuming feedback loop, resulting in degradation of sample quality and diversity—a phenomenon termed Model Autophagy Disorder (MAD). Conventional approaches recommend avoiding synthetic self-training to prevent such collapse.
This thesis provides a detailed investigation of MAD, combining theoretical analysis and empirical evidence to characterize its causes and effects. In addition, new methods are introduced for enabling generative models to self-improve using synthetic data while avoiding degradation by steering the models away from synthetic data distribution. The proposed techniques demonstrate that, with appropriate corrective mechanisms, self-generated data can be harnessed to enhance model performance beyond the current state-of-the-art. The results show that generative models can be made significantly more data-efficient and resilient, opening new directions for sustainable, self-improving generative learning systems
From Bayesian Nonparametrics to Symbols: Scalable and Accurate Model-Free Variable Selection for High-Dimensional Symbolic Regression
No full textThis thesis develops a suite of novel statistical frameworks and tools for scalable symbolic regression (SR) with a focus on high-dimensional regimes. SR seeks to discover closed-form mathematical expressions that explain the relationship between a response and a set of predictors, offering both interpretability and predictive accuracy. Despite its appeal, SR remains computationally challenging, particularly in large-p settings where the combinatorial explosion of model search space can render existing methods intractable.
The first chapter formulates SR as an ultra-high-dimensional Operator-Induced Structural linear regression problem. To navigate this vast model space efficiently, we introduce parametrics assisted by nonparametrics (PAN), an iterative framework utilizing nonparametric variable selection to enable scalable SR. We instantiate PAN as iBART, which alternates between Bayesian additive regression trees (BART)-based variable selection and feature synthesis. This iterative dimension reduction shrinks the search space to promising subspaces, significantly improving scalability and accuracy. Simulations demonstrate that iBART is reliable and efficient. In an application to single-atom catalysis, iBART identifies meaningful descriptors, offering insights into sintering-free catalyst design.
The second chapter addresses the scalability bottlenecks of existing SR algorithms in the large-p regime--a setting increasingly common in modern scientific applications. We propose PAN+SR, a two-stage framework that integrates ab initio nonparametric variable selection with any SR algorithm. We propose a novel clustering-based selection method operated on variable inclusion proportion ranks, which efficiently reduces dimensionality while minimizing false negatives, a key requirement for symbolic recovery. To evaluate PAN+SR in large-p settings, we design an SR benchmark comprising 35 real-world datasets and 100 synthetic datasets based on nonlinear equations in the Feynman Lectures on Physics. PAN+SR consistently improves the performance of 19 SR algorithms across diverse settings.
In the third chapter, we address two fundamental challenges in BART-based variable selection: high computational burden and unstable selection accuracy. We provide a comprehensive review of existing variable importance metrics and introduce a new measure based on variable count and rank statistics. Extensive numerical experiments show that the proposed measure consistently outperforms 7 existing BART-based methods across diverse settings. Its accuracy, robustness, and efficiency make it suitable for both recall-oriented screening and precision-focused selection.
Collectively, these contributions bridge nonparametric statistics and symbolic modeling, advancing the foundations of model-free variable selection and interpretable modeling in high-dimensional regimes
What Do New Teachers Need? How an RPP Worked to Tailor Professional Development for First-Time Teachers
No full textNumerical and Experimental Study of Single-Objective Lattice Light Sheet Microscopy for 3D Single-Molecule Super-Resolution Imaging of Cells
No full textFluorescence single-molecule localization microscopy has revolutionized the imaging of cellular architectures and molecular dynamics beyond the diffraction limit of light. Conventional illumination methods for fluorescence microscopy suffer from fluorescence background from out-of-focus emitters, which decreases the signal-to-background ratio of the image, and hence degrades localization precision for single-molecule imaging. Light sheet illumination, which optically sections the sample with a thin sheet of light, can mitigate this issue, and improves the performances of both diffraction-limited and single-molecule imaging. Moreover, lattice light sheet (LLS) microscopy, which implements a 2D optical lattice for light sheet illumination, has been shown to overcome some limitations of traditional light sheet microscopy systems. However, the quantitative analysis of LLS is incomplete. Additionally, in conventional LLS microscopy, the configuration of the excitation and detection objectives restricts the use of high numerical aperture objectives for illumination and detection, introduces a risk of sample contamination, and is incompatible with microfluidics.
In this thesis, I present numerical simulations of the propagation properties of LLSs in free space and in scattering environments, providing a quantitative perspective on their characteristics. Furthermore, I present a novel microscopy platform design for single-objective LLS microscopy, which addresses some of the limitations of the conventional dual-objective design of LLS setups. Additionally, we experimentally characterize the performance of the LLS in comparison to a Gaussian LS, and demonstrate that LLS illumination enhances both diffraction-limited and single-molecule super-resolution imaging over conventional illumination strategy. Finally, we implement point spread function engineering in our setup and apply it for single-particle tracking in 3D using LLS illumination. Our work introduces new designs for LLS microscopy, enhancing the applicability of single-objective light sheet setups in single-molecule localization microscopy. This advancement will enable the visualization of nanoscale cellular structures in 3D with improved localization precision and higher resolution, offering valuable insights into cellular biology and potential applications in various biological research fields
The Complexity of Verifying and Realizing Equilibria in Finite-Horizon Multi-Agent Systems
Full text linkSystems in which multiple independent agents interact are an increasingly common design pattern in computer science. Despite this, providing formal guarantees on the overall behavior of such systems is difficult due to the inherent complexity that arises when multiple agents interact. For this reason, there has been a growing focus on multi-agent systems in the field of computer science, as multi-agent systems provide a mathematical framework to analyze systems in which multiple agents with independent goals strategically interact with one another.
The majority of prior research primarily focused on infinite-horizon goals, which considered an entire infinite execution of a system. Following the development of finite-horizon temporal logics, more recent work has shifted to finite-horizon goals, which can be completed in a finite number of steps and offer better computational properties. Furthermore, finite-horizon goals provide a better model for notions like ``completion," which are fundamental to the study of real computer systems. Therefore, recent work in multi-agent systems has trended towards analyzing systems where agents have finite-horizon goals. The natural next step is to consider how such systems should be analyzed, which has led researchers to study equilibria in these systems.
Equilibria are key concepts from the field of game theory that mathematically define stable group behaviors that emerge from repeated interactions in multi-agent systems. The most common way to analyze equilibria is through the lens of the verification and realizability decision problems. Given an equilibrium concept, the verification problem asks whether an input set of behaviors satisfies the equilibrium condition, whereas the realizability problem takes an equilibrium concept as input and asks whether there exists a set of behaviors that satisfies the equilibrium concept.
This dissertation studies the computational complexity of the verification and realizability decision problems for equilibria in finite-horizon multi-agent systems. Although there have been many recent works in this area, this dissertation advances the literature in three key ways. First, it considers how different components of the system affect its computational complexity. In a system in which each agent has a goal, both the specification of the system and the specification of the agents' goals influence the overall complexity of analyzing the multi-agent system. Second, it explicitly considers the construction of the transition function, which describes how executions of the system unfold based on the agents' choice. Since the transition function is the most complex part of the specification of the system, it has the greatest impact on the system's complexity-theoretic properties. By explicitly considering the size of the goal specification and the transition function (and, therefore, the size of the system), this dissertation offers a more fine-grained and complete consideration of problems that have been analyzed previously in recent works. Finally, it extends this analysis to probabilistic systems and addresses issues of numerical representation, a key point of consideration that is largely missing in the literature.
There are several highlights in the results presented in this dissertation. First, the results in this dissertation resolve the long-standing and actively worked-on open problem of non-deterministic finite word automaton realizability. Second, the automata-based techniques developed for qualitative settings in this dissertation are extended to quantitative systems through the use of ``satisficing" goals, which, when compared to the often-used optimization goals in quantitative systems, offer benefits from both a computational and conceptual lens. Finally, for probabilistic systems, our analysis of the verification problem establishes a counterintuitive relationship between the two most popular equilibrium concepts used in the literature, thereby challenging the conventional understanding of these fundamental concepts
How African Immigrants Interpret The Connection Between Their Religion and Health.
Full text linkReligion can positively and negatively influence individuals’ health behaviors. While religion can deter risky behaviors like alcohol abuse, it can sometimes discourage seeking healthcare. Religion has primarily been presented as a barrier to seeking healthcare. Additionally, African immigrants in the United States of America have received less coverage in research about their religion and health despite being part of a demographic group (Blacks) that has developed a mistrust of the medical health system in the U.S. due to historical treatment.
This thesis examines the health experiences of African immigrants in Houston, Texas, focusing on how they interpret the connection between their religion and physical health. It also explores the perceived role that religious congregations play in the health experiences of African immigrants. Drawing on in-depth interviews of 37 Christian African immigrants living in Houston, I find that religion acts as a pathway to healthy living and seeking healthcare among African immigrants. Thus, religion provides a framework for a positive perspective on medical healthcare. By focusing on African immigrants, this study serves as a case for understanding the health experience and behaviors of highly educated and religious populations