15064 research outputs found
Sort by
Digital support systems to improve child development in Peru: A cluster-randomized controlled open-label trial
No full textDigital technologies have the potential to transform early childhood development (ECD) interventions by delivering personalized support at scale. We conducted a cluster-randomized controlled trial in rural Peru to evaluate the effectiveness and cost-effectiveness of an artificial intelligence–supported digital parenting chatbot as well as traditional home visits as interventions to improve child development. Among 2461 caregiver-child dyads, both the digital and home-visiting interventions improved child development outcomes at 2.5 years of age, with standardized effect sizes of 0.11 and 0.17, respectively. At 1/15 of the cost of in-person support, the digital intervention yielded superior cost-effectiveness. These findings suggest that digital platforms can be a viable, scalable alternative to support children’s development in resource-constrained settings.</p
The Definition of Happiness: Topic Modeling and Retrieval-Augmented Generation for Historians of the Eastern Miscellany.
No full textThis thesis explores the cultural construction of selfhood in Republican-era China through the lens of shenjing shuairuo 神經衰弱, or neurasthenia. By analyzing a corpus from The Eastern Miscellany (Dongfang zhazhi 東方雜誌), using both close and distant reading methods, this study examines how they can supplement each other’s limitations. Furthermore, it shows how historians can leverage the power of large language models to study their corpus. The research highlights the term’s ambiguous usage and how it shaped modern Chinese personhood and mental health
Mast cells promote pathology and susceptibility in tuberculosis
No full textTuberculosis (TB), caused by the bacterium Mycobacterium tuberculosis (Mtb), infects approximately one-fourth of the world’s population. We reported an increased accumulation of mast cells (MCs) in the lungs of macaques with active pulmonary TB (PTB), compared with those with latent TB infection (LTBI). MCs respond in vitro to Mtb exposure via degranulation and by inducing proinflammatory cytokines. In the current study, we demonstrate an increased production of chymase by MCs in granulomas of humans and macaques with PTB. Single-cell (sc) RNA sequencing analysis revealed distinct MC transcriptional programs between LTBI and PTB, with PTB-associated MCs enriched in interferon gamma, oxidative phosphorylation, and MYC signaling. In a mouse model, MC deficiency led to improved control of Mtb infection that coincided with reduced accumulation of lung myeloid cells and diminished lung inflammation at chronic stages of infection. Airway transfer of MCs into wild-type Mtb-infected mice showed increased neutrophils, decreased recruited macrophages, and elevated Mtb dissemination to the spleen. Together, these findings highlight MCs as active drivers of TB pathogenesis and potential targets for host-directed therapies for TB
Optimization and Control for Scalable Modeling and Inference in Dynamical Systems
No full textIn this thesis, we investigate methodologies for discovery, assimilation, and control of (stochastic) dynamical systems, and introduce algorithms that exploit structural properties for efficiency. Data-driven modeling problems in physics and engineering are governed by high-dimensional dynamical systems, where direct simulation and control can become prohibitively expensive. Chapter 2 begins with the discovery of mathematical descriptors directly from data, focusing on nonautonomous and translation-invariant systems. By combining the Lagrangian frame of reference with locally time-invariant Koopman approximations, we obtain low-rank linear time-varying surrogate models with a priori predictive error estimators. We validate the approach on canonical flow and 2D advective transport problems. In Chapter 3, we shift focus to stochastic systems, and address uncertainty quantification in multiscale models where stiffness impedes statistical convergence. We propose reduced-order probability density equations, which employ closure terms defined as state-space conditional expectations and learned directly through data assimilation strategies, such as nudging. Chapter 4 generalizes this methodology to high-dimensional observables. The methodology is then applied to stochastic power system models to design estimators for extreme-event probabilities, such as cascading line failures. Numerical results demonstrate both stability and efficiency, outperforming naive Monte Carlo simulation and kernel density estimations. In Chapter 5, we address the problem of control, which is a central problem in the characterization of extreme events. We present an optimize-then-discretize framework for linear-quadratic control governed by time-inhomogeneous dynamics. Our method employs a modified overlapping Schwarz decomposition in continuous-time, which ensures convergence through the exponential decay of sensitivity property in Hamiltonian systems. Unlike discretize-then-optimize approaches, the framework allows flexible numerical integration schemes while preserving stability. This contribution paves way for distributed nonlinear control as well as deep learning applications, whose analysis is left for future work
Unsaying the Two Truths: A Translation, Critical Edition, and Introduction to Two Madhyamaka Texts by Karmapa VIII Mikyö Dorje (1507-1554)
No full textThis dissertation is comprised of a translation, critical edition, and introduction to two Madhyamaka (‘Middle Way’) philosophical works by Karmapa VIII, Mikyö Dorje (Karma pa Mi bskyod rdo rje, 1507-1554), hierarch of the Tibetan Buddhist Karma Kagyü (Karma Bka’ brgyud) monastic order. Those works, whose main polemical target is Tsongkhapa Lozang Drakpa (Tsong kha pa Blo bzang grags pa, 1357-1419), progenitor of the Dalai Lama’s Gelukpa (Dge lugs pa) order, present worldly reality as being so tainted by erroneous conceptuality that it is unknowable, with Buddhist enlightenment being the complete cessation of all appearances since they are merely the product of delusion. This view is in stark contrast to the metaphysics and epistemology of Tsongkhapa, who considers phenomena to be knowable conventionally even if, ultimately, they lack any intrinsic nature. As such, the debate implicates such questions as what we can know for sure about the world, how this knowledge (or lack thereof) grounds our ethical decision making, and even how we establish hierarchies of authority. The Madhyamaka texts on which my dissertation is focused, The Praise to Dependent Arising (Rten 'brel bstod pa) together with its autocommentary and The Glorious Song of Delight on the Unerring Single Way (Tshul tshul gcig pa’i nges don ’khrul bral gyi glu dpal dbyangs can dga’ ba), have not previously been translated or studied academically and were likely among the large tranche of texts banned by the government of Dalai Lama V, Ngawang Lozang Gyatso (Ngag dbang blo bzang rgya mtsho, 1617-1682). As I argue in my introduction to the texts, Mikyö Dorje follows a hermeneutically conservative tradition that takes a philosophically radical approach to reality, while Tsongkhapa’s less conservative interpretation of Buddhist scripture follows a strain of moderate realism. While mainstream narratives portray Tsongkhapa’s view as the “central philosophy of Tibet,” the force of Mikyö Dorje’s compelling critique of that view, which largely went unanswered by scholars of the Gelukpa hegemony, suggests Tsongkhapa was more a politico-historical winner than a philosophical one. The dissertation thus centers one of the most important yet under-studied interpretations of Madhyamaka, offering a set of translations, critical editions, and intellectual-historical context in addition to explaining the motivations and justifications of Mikyö Dorje’s liberation-oriented Buddhist philosophical view
Shock Acceleration in the Intracluster Medium: Implications of Micromirror Confinement
No full textMerging galaxy clusters exhibit strong observational evidence for efficient particle acceleration in the intracluster medium (ICM), particularly in the form of synchrotron-emitting radio relics and halos. Cosmic-ray (CR) electrons are likely accelerated (or reaccelerated) at merger and accretion shocks via diffusive shock acceleration. However, in the presence of the large diffusion coefficients, one would naively expect in the rarefied, relatively unmagnetized ICM, this acceleration—in particular, the maximum proton energy ( E max )—is limited by long acceleration times. On the other hand, recent work on CR transport suggests that the diffusion coefficient can be suppressed in ICM-like environments. In this picture, deviations from local thermodynamic equilibrium can trigger the mirror instability, creating plasma-scale magnetic structures, or “micromirrors,” that efficiently scatter CRs. In this paper, we investigate the implications of micromirror confinement for shock acceleration in the ICM. We demonstrate that micromirrors enforce a minimum value of E max ≳ 100 GeV that does not rely on CR-driven magnetic field amplification. We also discuss micromirror confinement in the context of cosmological simulations and γ -ray observations, and present a simulation of a Coma-like merging cluster that self-consistently includes CR acceleration at shocks, with an effective diffusion coefficient set by micromirrors. We show that the introduction of micromirrors yields simulated galaxy clusters that remain consistent with γ -ray observations
The Invisible Threat of TikTok Against US National Security: Chinese State-Affiliated Intelligence Collection and Cyber Enabled Influence Operations
No full textThere is a general misunderstanding on why the US Government has pressured TikTok, a social media company owned by the Chinese internet company ByteDance,1 to divest the US version of its application. Although TikTok and the US national security community attempted to develop an alternative solution, the US Government still cites insurmountable security concerns of Chinese government influence – Chinese state-owned enterprises maintain a 1% stake in ByteDance domestic operations and the CCP ensures access to collected data through its various national legal requirements.2 In order to illuminate TikTok’s threat to US national security, I will analyze various open-source reports, declassified intelligence reports, court documents derived from FBI special investigations and national security assessments, translated Chinese source documents, and media reports. I will provide a historical background on TikTok and China’s development of cyber capabilities. Then, I will define how TikTok acts uniquely as a Chinese data collection platform by assessing its privacy policy, data harvesting techniques, addictive algorithms, and Chinese government accessibility to these data sets. I will explore the implications of bulk data along with artificial intelligence (AI), to include the advantages of processing, targeting, and self-learning models. I will also focus on the covertness of TikTok intelligence collection, influence operations, and its over-arching objectives. Then, I plan to analyze People’s Liberation Army military doctrine on civil-military fusion and joint information operations to predict potential militarization of TikTok intelligence and influence capabilities through information warfare. Finally, I will conclude on summarizing the TikTok threat to national security and develop feasible policy options to mitigate this risk
Inverse Statistical Physics: Connections to Machine Learning and Applications in Biology
No full textStatistical mechanics was historically significant for its ability to link the microscopic descriptions of matter and energy to the macroscopic observations of thermodynamics. Recent cross-disciplinary work has utilized insights from statistical mechanics to solve the inverse problem---going from observable phenomenon to underlying interactions and principles---and has done so from applications in protein research to theory in neuroscience. Inverse statistical physics relies on a key property of information entropy: that fitting a distribution to data such that it obeys given observed constraints (but is otherwise as maximally entropic as possible) leads to a probabilistic model of the system that is least biased from unobserved assumptions, which leads to maximal predictive power. These maximum entropy models belong to a wider class of regularly used architectures in machine learning, known as energy-based models. When such models are fit to real data of complex multi-dimensional systems, ideally the learned distribution is able to generate states exemplary of the ground truth. In practice, however, this is often not the case; specialized sampling must be performed to generate desired outputs. For example, energy-based models of sequences of evolutionary related families of proteins have the ability to learn the generic constraints necessary to make novel functional sequences, which have been validated by in vivo experiments. However, these learned energy functions require re-scaling by a temperature parameter in order to sample novel functional sequences. Here we utilize minimal and physically motivated energy-based models in order to systematically interrogate the differences between the data-generation processes of ground truth and learned models sampled at varying temperatures. This lends itself well to an examination of the surprising ability of temperature tuning of learned energy functions---a poorly understood heuristic used across machine learning---to improve sampling performance. Whether the post-hoc sampling temperature need be raised or lowered, and by how much, depends on several factors: choice of objective function, amount of training data, and most importantly, properties of order and disorder inherent to the true system. Crucially, we show that the need to lower temperature to improve generative performance arises from a tendency of fit models to overestimate the probability mass on excited states when the number of training data is low and the ground truth is characterized by a strong preference for producing few ground states---induced by large "energy gaps" or low ground truth "temperature." Additionally, we show via a minimal setting that the temperature tuning phenomenon may be directly linked to a wide array of empirical evidence for a synergistic cluster of amino acids, or sector, within a sequence that is responsible for determining the functionality of that protein sequence
Judeo-Arabic and Hebrew as Competing Languages in the Mamluk Period
No full textThis article examines the linguistic choices of Jewish communities in Egypt, Palestine, and Syria during the Mamluk period, focusing on two types of texts: letters and legal documents. Jewish writers navigated a tension between their fluency in Arabic and their obligation to preserve Hebrew as the language representing their distinct religion and culture. The article analyzes the various ways Jewish writers grappled with this tension
Molecular Engineering of Remote-gate Field-effect Transistor Sensors for Ion Monitoring in Water
No full textReliable, real-time ion monitoring is essential for understanding and managing water quality, but gold-standard analytical methods (e.g., inductively coupled plasma and ion chromatography) remain laboratory-bound and poorly suited for continuous field deployment. Meanwhile, field-effect transistor (FET) sensors offer fast, sensitive, and low-power electrical readouts, but conventional back-gate and solution-gated configurations often suffer from drift, hysteresis, and device-to-device variability—especially when low-dimensional materials serve simultaneously as the sensing interface and the electronic transducer. This dissertation develops and validates a molecular-engineering framework for remote-gate field-effect transistor (RGFET) sensors that decouple the solution-facing interface from the semiconductor transducer, enabling stable, reproducible sensing of heavy metal and nutrient ions in water with varying pH values. Chapter 1 presents the background of ions in water environment, summarizes the current state of ion detection technologies, and introduces FET water sensors for ion detection. In Chapter 2, the dissertation first establishes how remote-gate structure suppresses non-ideal behaviors that commonly limit aqueous FET sensing. Using reduced graphene oxide (rGO) confined to the solution interface and capacitively coupled to a commercial metal-oxide-semiconductor field-effect transistor (MOSFET), the platform prevents current flow through the sensing film and mitigates interfacial redox processes, trapped-charge effects, and material instabilities that drive drift and hysteresis. A systematic mechanistic study links rGO sensing performance to film thickness/coverage and reduction conditions, and introduces intrinsic electrochemical metrics extracted from transfer characteristics to quantify stability. With an optimized multilayer rGO interface, the RGFET achieves near-Nernstian pH sensing with high linearity, minimal drift, and negligible hysteresis, while maintaining high device yield and reproducibility. Building on this foundation, Chapter 3 demonstrates molecular engineering of device interfaces to tune rGO electrochemical properties and enhance ion sensing. Self-assembled interlayers of (3-aminopropyl) trimethoxysilane (APTMS) and hexamethyldisilazane (HMDS) are used to control surface energy and selectively bias the chemical composition and hydrophilicity of deposited rGO, thereby modulating proton sensitivity and interfacial stability. In parallel, linker chemistry based on pyrene derivatives is quantitatively characterized using the RGFET, enabling estimation of surface charge density and linker surface density resulting from functionalization. By leveraging these insights to increase probe density on the sensing surface, the platform achieves improved heavy-metal response—illustrated by enhanced Pb²⁺ sensitivity using glutathione capture chemistry enabled by optimized linker/substrate interface. Finally, the dissertation translates these principles into deployable systems for practical water monitoring. In Chapter 4, a fully portable and reusable Pb²⁺ sensor is realized using a sensor printed circuit board (PCB) bearing a graphene film remote gate, a compact analyzer PCB with on-board electronics and wireless communication, and a smartphone interface for real-time display and threshold-based warnings. The portable sensor exhibits a high sensitivity of 21.7 % when detecting its LOD value of 1 nM (~ 0.2 ppb), while the sensor PCB provides adequate adhesion to the graphene ink, allowing facile deposition and removal, which results in the ability to reuse the sensor PCB repeatedly. In this way, the system detects sub-ppb lead rapidly (on the order of a minute) with strong selectivity and a regenerable sensing surface. To realize nutrient ion monitoring, a multiplexed RGFET array is presented in Chapter 5 for simultaneous nitrate, nitrite, and phosphate detection using ion-selective remote-gate modules. The array exhibits near-Nernstian potentiometric behavior (54.2, 48.7 and 40.5 mV/decade respectively) across environment-relevant concentration ranges (10-2–10-5 M). A machine-learning analysis model is integrated to resolve cross-sensitivity in mixed-ion environments, achieving prediction accuracies with R2 values exceeding 0.98 for all nutrient ions. The array exhibits near-Nernstian potentiometric behavior across relevant concentration ranges and supports continuous monitoring in a flow-cell configuration under dynamic conditions. Collectively, these advances position remote-gate FET sensing—paired with molecular engineering, portable electronics, and data-driven analysis—as a scalable pathway toward robust, real-time ion monitoring for environmental and water-quality applications. Intellectually, this dissertation establishes a mechanistic framework that connects molecular engineering and materials chemistry to sensor accuracy, stability, and key performance metrics, yielding general design principles for reliable FET-based water sensors. Broadly, the resulting portable and connected platforms enable low-cost, field-deployable monitoring of toxic metals and nutrients, shifting ion analysis from episodic laboratory testing to continuous, real-time monitoring that support faster response and wider access