545298 research outputs found
Sort by
Using Multimodal AI to Predict Lung Adenocarcinoma Recurrence
Get PDFLung adenocarcinoma (LUAD), a subtype of non-small cell lung cancer (NSCLC), is characterized by significant histologic and clinical heterogeneity. Although early-stage NSCLC patients are commonly treated with curative resection, a substantial proportion experience recurrence within five years. Current prognostic tools such as tumor, node, and metastasis staging fail to fully capture tumor aggressiveness for patients with different characteristics. This underscores the need for more accurate, LUAD-specific risk stratification approaches. Recent advances in artificial intelligence (AI) and machine learning (ML) offer new opportunities to derive prognostic biomarkers from routinely acquired data such as radiological scans and histopathology tissue slides. However, existing pathomic models often overlook the spatial context within the tissue slides, and most of them do not explicitly consider different NSCLC subtypes. Furthermore, most AI/ML models rely on unimodal data, missing the complementary insights available from various data modalities. This dissertation addresses these limitations by developing novel AI/ML-based methods to predict cancer aggressiveness in early-stage LUAD patients, using various data modalities such as hematoxylin & eosin-stained whole slide images, preoperative computed tomography scans, and clinical information. The proposed framework captures tumor heterogeneity through spatially-aware pathomic features and integrates different data modalities to improve risk stratification. In doing so, this work aims to improve personalized risk prediction and inform adjuvant therapy decisions by adding additional evidence on top of the standard clinical assessments
Tracking Microbial Water Quality with Advancing Techniques: Remote Sensing and Next Generation Sequencing
No full textMicrobial water quality monitoring is critical for protecting public health and environmental sustainability, yet traditional methods face limitations in spatial coverage and efficiency. This dissertation explores the integration of remote sensing and next-generation sequencing techniques to enhance microbial water quality surveillance across diverse aquatic environments.The first study evaluated remote sensing for turbidity monitoring in Los Angeles coastal waters from July 2021 to March 2024. Correlation analysis revealed a positive relationship between satellite-derived and field-measured turbidity, with machine learning models achieving improved predictive accuracy. Seasonal trends showed elevated turbidity during wet months due to stormwater runoff, demonstrating the potential for large-scale and efficient turbidity monitoring.
The second study investigated the feasibility of using satellite observations to predict fecal indicator bacteria levels in coastal waters. A moderate correlation was observed between Sentinel-2 derived total suspended matter and in situ Escherichia coli concentrations, validated using historical California Water Board datasets. Results indicate satellite data can effectively estimate E. coli concentrations, enhancing pollution event warning systems for public health protection.
The third study evaluated antibiotic resistance genes (ARGs) and microbial source tracking markers in surface waters adjacent to concentrated animal feeding operations (CAFOs) in Michigan. Twenty samples were collected from impacted areas and reference sites across two seasonal periods and analyzed using culture-dependent methods, quantitative PCR, and shotgun metagenomic sequencing. Results demonstrated elevated concentrations of E. coli and antibiotic-resistant bacteria in near CAFO areas. Sites impacted by agricultural activities exhibited significantly higher diversity of ARGs and greater bacterial community richness compared to reference locations. When compared to a previous study, tetracycline-related resistance parameters exhibited a more pronounced antibiotic resistance trend relative to other antibiotic classes, suggesting that tetracycline resistance may serve as a particularly sensitive indicator of CAFO influence on environmental microbial communities.
This research demonstrates that integrating remote sensing with machine learning and applying comprehensive molecular approaches can significantly advance microbial water quality monitoring. The methodologies explored in this dissertation offer water quality managers and public health officials improved approaches for proactive monitoring, early warning systems, and evidence-based decision-making to protect both environmental health and human welfare in diverse aquatic environments
Stochastic Quantum Algorithms for Quantum Simulation
No full textThis dissertation introduces several quantum algorithms aimed at accelerating the realization of useful simulations on quantum computers.In Chapter 1, we introduce a hybrid quantum-classical approach for simulating a class of open system dynamics called random-unitary channels. These channels naturally decompose into a convex combination of unitary evolutions, which can be efficiently sampled and run as independent, ancilla-free circuits. We implement simulations of open quantum systems up to dozens of qubits and with large channel ranks on IBM hardware.
In Chapter 2, we extend this stochastic approach to general quantum channels, which we simulate using ensembles of low-depth, single-ancilla circuits. We demonstrate the efficiency of this method by preparing damped, many-qubit GHZ states on IBM hardware. The technique further inspires two Hamiltonian simulation algorithms with asymptotic independence of the spectral precision, reducing resource requirements by several orders of magnitude for a benchmark system.
In Chapter 3, we introduce stochastic Zassenhaus expansions (SZEs), a class of ancilla-free quantum algorithms for Hamiltonian simulation. These algorithms map nested Zassenhaus formulas onto quantum gates and then employ randomized sampling to minimize circuit depths. Unlike Suzuki-Trotter product formulas, which grow exponentially long with approximation order, the nested commutator structures of SZEs enable high-order formulas for many systems of interest. For a 10-qubit transverse-field Ising model, we construct an 11th-order SZE with 42x fewer CNOTs than the standard 10th-order product formula. Further, we empirically demonstrate regimes where SZEs reduce simulation errors by many orders of magnitude.
In Chapter 4, we propose a dissipative algorithm to prepare the Gibbs state of commuting Hamiltonians, with extensions to general systems. The algorithm prepares the Gibbs state of a spanning tree subgraph of a Hamiltonian in linear time. It then probabilistically implements the remaining interactions as perturbations to this tree. The circuit depth scales linearly in the total number of interactions, making it amenable to near-term applications. For low temperatures, the runtime scales exponentially with the frustration of the ground state, enabling the efficient simulation of a broad class of low-frustration systems. In particular, this gives a linear-time quantum algorithm for finding the ground state of commuting, frustration-free Hamiltonians
Optimal Transport in Stochastic and Deterministic Settings with Applications to PDEs
Get PDFThis dissertation analyzes nonlinear partial differential equations (PDEs) arising from transport phenomena in both deterministic and stochastic settings. Central to the analysis is the use of Lagrangian and particle-based perspectives. These perspectives appear in several forms throughout the dissertation. One setting involves the analysis of a density-constrained continuity equation with drift, where the Lagrangian flow associated with the drift field is used to derive new geometric properties of the saturated region. Another uses an interacting particle system, arising from a stochastic optimal transport problem, to analyze the nonlocal Stefan problems. A third appears in the development and analysis of an accelerated Lagrangian scheme for gradient flows in the Wasserstein space of probability measures.In Chapter 2, we extend the classical Hele-Shaw model for tumor growth by allowing the maximum cell density (height function) to vary in space and time. Under standard assumptions on the initial data and smoothness assumptions of the source, drift, and height functions, we establish well-posedness of the resulting system. Additionally, under a congestion condition, we derive an Aronson–Benilan-type lower bound on the time derivative of the pressure. This estimate implies that the tumor region and normalized density are non-decreasing in time along the Lagrangian flow associated with the drift velocity field. This work has led to the publication [Chu23].In Chapter 3, we investigate a non-local Stefan problem modeling melting and freezing phenomena with long-range interactions. Weak solutions are constructed using a density constrained stochastic optimal transport problem with a strictly monotone cost, where particles evolve according to a Levy process with long-range jumps. For the stable melting problem, we prove exponential in time convergence in the L1-norm of the enthalpy variable to the optimal target measure. In the unstable freezing case, the long-range jumps enable the construction of weak solutions with a prescribed initial temperature profile and any sufficiently large terminal positivity set. Furthermore, when both the initial temperature profile and terminal positivity region are specified, we establish uniqueness of weak solutions to the freezing problem by identifying weak solutions with optimizers to our underlying stochastic optimal transport problem. This work has led to the publication [CKKN25].In Chapter 4, we introduce an accelerated implicit Lagrangian scheme for Wasserstein gradient flows. Although accelerated schemes are designed to improve upon first-order methods, previous accelerated schemes have only been able to rigorously establish, at best, an O(√τ) convergence rate under general assumptions, which is worse than the O(τ) rate typically achieved by first-order schemes, where τ denotes the step size. In contrast, our scheme recovers the O(τ) convergence rate under general conditions, specifically, when the energy is λ-displacement convex and Wasserstein differentiable, ensuring that it performs at least as well as first-order methods. In addition, when the energy functional is sufficiently smooth, our scheme further achieves the optimal O(τ²) convergence rate in the 2-Wasserstein metric, through a calculus-based argument that leverages the differential structure of Wasserstein space.Beyond convergence rates, our scheme satisfies favorable stability properties. Along the discrete solutions, the energy is nearly decreasing in general. When the energy is λdisplacement convex with λ ≥ 0, the L2-norm of the Wasserstein gradient is non-increasing over time. For λ < 0, this norm grows at most exponentially, with an asymptotically optimal rate. Furthermore, when the energy is both λ-convex with λ ≥ 0 and L-smooth, the energy becomes non-increasing, and the L2-norm of the gradient decays exponentially in time at an asymptotically optimal rate
Modulating the Electronic Transport of 2D Sb2Te3 Nanoplates by Coinage Metal Intercalation
No full textThermoelectric materials are particularly relevant to the current energy infrastructure and demands of the 21st century, converting waste heat into usable electricity. The solution intercalation of zerovalent copper into Sb2Te3 nanoplates, a well-established thermoelectric material, is reported. The copper intercalant is homogeneously distributed throughout the nanoplates, confirmed by scanning transmission electron microscopy coupled with energy-dispersive X-ray spectroscopy. The copper composition was shown to be 6 at. % by X-ray photoelectron spectroscopy. Copper ordering within the van der Waals gaps of the nanoplates is confirmed by selected area electron diffraction. Fabrication and thermoelectric property measurements of single-crystal Sb2Te3 and Cu-Sb2Te3 nanoplate devices show effective modulation of electrical conductivity and Seebeck coefficient with Cu intercalation. X-ray photoelectron spectroscopic studies in the valence-band region reveal additional electronic states from copper that appear near the Fermi energy, postulated to act as electron acceptors, leading to modulation of the electronic transport properties
Energetic Chemistry in the Atmospheres of Brown Dwarfs
Get PDFBrown dwarfs are a unique laboratory for understanding atmospheric physics in planetary-mass objects. Though they form similarly to stars, brown dwarfs have cooler temperatures and lower masses, making their atmospheres often more analogous to those of giant planets. While significant progress has been made in characterizing their bulk properties and lower atmospheres, the upper atmospheres of brown dwarfs remain largely unexplored. Upper atmospheric regions are critical because they mediate interactions between the space environment, atmosphere, and planetary interior, absorbing high-energy radiation externally, while internally coupling with magnetic fields and exchanging energy with the lower atmosphere through radiative transfer and atmospheric waves.This dissertation investigates the properties and evolution of brown dwarf atmospheres, with a focus on energetic chemistry in the upper atmosphere, using a combination of new observations, theoretical modeling, and instrumentation development. I begin with new observations of the benchmark brown dwarf HD 33632 Ab, examining the importance of clouds, disequilibrium chemistry, and composition in comparing near-infrared spectra with evolutionary models. Next, I present the first high-resolution spectroscopic search for upper atmospheric H3+ emission in the near-infrared L band, placing upper limits on its emission and probing potential differences in auroral chemistry between brown dwarfs and giant planets. I then model the photochemical and thermal response of brown dwarf upper atmospheres to energetic flares from active M dwarf companions, predicting observable spectral signatures. Finally, I describe image simulations in support of the upcoming High-resolution Infrared Spectrograph for Exoplanet Characterization (HISPEC) instrument at Keck Observatory, which will enable next-generation high-resolution studies of exoplanet and brown dwarf atmospheres. Collectively, these studies underscore the complexity and current gaps in our understanding of brown dwarf upper atmospheres and help lay the foundation for future advances in the field
Plots on the Periphery: Black Women-Led Grassroots Planning in Inland Southern California
Get PDFThis dissertation traces an overlooked grassroots planning tradition, led by Black women, to counter racial segregation, exclusion, and discrimination in an understudied setting: the urban periphery. The Antelope Valley and the Inland Empire, two inland Southern California areas long shaped by white spatial imaginaries, have nonetheless been home to Black households for generations and are increasingly so due to ongoing Black suburbanization and displacement. In this study, I analyze how Black people worked collectively to transform these landscapes of entrenched racial inequality. Through a conjunctural analysis using qualitative research methods, I foreground three seismic moments in which Black women impactfully mobilized their communities to disrupt the existing spatial order, by planning placemaking and policy interventions affirmative of Black life. In the first, club women initiated the midcentury development of a park in Sun Village, a Black town in the Antelope Valley. In the second, following the Watts Rebellion in 1965, Black mothers mobilized hundreds of Riverside and San Bernardino residents to desegregate their de facto segregated public schools, by organizing boycotts, opening Freedom Schools, and advocating for busing and open enrollment policies. In the third, mothers in the Antelope Valley are leading anti-carceral efforts to reduce police presence in public spaces, particularly schools. I find that particular social, cultural, historical, political, and economic forces interacted to form conjunctures in which Black women effectively built counterhegemonic movements in urban peripheries, places distinctly marked by inequitable development and a relative dearth of social safety nets. Among these forces are: the emergence of replicable organizing models and programs, and rising political will for direct action amongst the public and for progressive change by government officials, during the Second Reconstruction and Third Reconstruction; and, contradictory development trajectories creating a paradox of Black economic opportunity and residential exclusion. Within this grassroots planning tradition, I find that Black women’s use of rival geographies to organize their communities reflect spatial practices of fugitivity, including marronage, garreting, and plot-making. Other significant characteristics of this tradition include a capacious ethic of collective care and the feminization of socially reproductive labor
Gendering Justice in the Chilean Legal System: Legal Actors’ Perspectives and Unequal Access to Justice
No full textWithin the last ten years, a number of Latin American legal institutions have designed strategies to promote a “gendered perspective” (perspectiva de género)— i.e. a method to detect gendered inequalities in the legal process— in legal institutions to improve access to justice for underrepresented groups, particularly women and the LGBTQIA+ community. Thus, this project asks: To what extent are the Chilean Supreme Court’s and Prosecution Offices’ recently implemented “gender-sensitive” approaches transforming the Chilean criminal justice system, and what are the experiences and perceptions about justice of those victimized by gendered crimes? Using ethnographic methods and drawing from legal anthropology, socio-legal studies and legal feminism, this research reveals that a gendered perspective entails a new “epistemic culture” (Knorr-Cetina 1999) in the Chilean legal system as it enshrines a new “way of seeing” (Goodwin 1994). The process of attempting to promote gender equity in the legal system produces a series of “epistemic frictions” (Kruse 2016) as actors who incorporate a gendered perspective in their everyday practices interact with those have not yet embraced this “new way of seeing,” or outright reject it. This project draws from seventeen months of fieldwork in Chile, in which I conducted court observations, participant observation, document analysis, and open-ended interviews with judges, prosecutors, public defenders, activists, and users of the court, in order to interrogate these questions
Low Complexity Coding and Learning Techniques for Resilient Millimeter-Wave Networks
No full textMillimeter-wave (mmWave) communication is a promising technology for meeting the high data rate, low latency, and high reliability demands of data-intensive applications such as virtual reality, cloud gaming, and autonomous systems. Despite its potential, mmWave communication faces significant challenges including sensitivity to link blockages, interference to passive users, limited theoretical understanding of fundamental performance limits, and the computational complexity of evaluating these limits in large networks. This dissertation addresses these challenges through low-complexity techniques and analyses in four main parts: Multilevel Coding, Hybrid Scheduling, Coexistence with Passive Users, and Gomory-Hu Trees. In the first part, we develop Multilevel Diversity Coding (MDC) schemes, which are proactive mechanisms that offer resilience without prior knowledge of blockages. Multilevel codes accommodate distinct Quality of Service (QoS) requirements of various information streams with different priorities. Despite their advantages, these codes may have high design and operational complexity. To address this challenge, this dissertation proposes low-complexity MDC schemes. First, we develop symmetric MDC schemes that encode information streams across space and time. We introduce an optimization framework to efficiently select the design parameters such that distinct QoS requirements are accommodated, while suitably balancing the trade-off between average rate and graceful performance degradation. Additionally, we develop a lower-complexity MDC scheme that approximates well the aforementioned trade-off. Our evaluations, carried out also within ns-3, show that these codes lead to favorable trade-offs between rate, delay, and outage probability. Extending this, we develop a low-complexity asymmetric MDC scheme that leverages unequal blockage probabilities to further improve reliability. We characterize its achievable rate region and derive capacity regions of certain scenarios. We show that our scheme is information-theoretically optimal when there are two priority levels across information streams. In the second part of the dissertation, we develop a hybrid scheduling mechanism that determines which network paths to use and how to schedule them to achieve a desired end-to-end packet rate in the presence of link blockages and channel variations. Towards this goal, we first develop proactive transmission mechanisms to build resilience in advance, while achieving a high packet rate. Building on this, we design an efficient path selection algorithm and integrate it with our deep reinforcement learning approach. Our evaluations show that the hybrid mechanism enables agile, decentralized adaptation to network dynamics in realistic environments. The third part of the dissertation focuses on mitigating interference to passive users. The mmWave spectrum is shared by both active users and passive users, where the latter do not transmit and are therefore difficult to detect, yet can be significantly affected by interference. We develop a method that limits the interference to passive users with a small penalty to the throughput of active users. We formulate a linear program, derive lower bounds on the achievable rates of active users, and characterize the number of required paths to achieve a target rate while limiting the interference. Finally, we establish a connection to the problem of (information theoretically) secure communication over mmWave networks. In the final part of the dissertation, we focus on the efficient evaluation of performance limits and show that the well-known Gomory-Hu algorithm can be used to efficiently compute information-theoretic rate characterizations, such as capacity for every pair of nodes in the network
Survivals of Pharaonic Religious Practices in Contemporary Coptic Christianity, version 2
Get PDFThe concept of “survivals” has provoked heated discussions among scholars of various disciplines within the humanities and the social sciences. In the case of Egypt the polemics have been most vehement between those who trace contemporary popular beliefs and practices back to Pharaonic times and others who reject the idea altogether. The perspectives of “analogy,” “continuity and change,” and “living traditions” have opened the way to alternative approaches to the subject. Urbanization and globalization have profoundly changed Egyptian culture and prompted the abandonment of most religious practices belonging to the Egyptian lore. However, some aspects of Pharaonic religious practices can still be observed in Coptic Christianity. These practices are tied to the Coptic calendar, funerary rituals, visits to the dead, and mulids