150813 research outputs found
Sort by
Coherent photoproduction of ρ0, ω and excited vector mesons in ultraperipheral PbPb collisions
No full textThe invariant-mass distribution for the coherent photoproduction of dipions in ultraperipheral PbPb collisions is measured using data, corresponding to an integrated luminosity of 224.6 ± 9.6μb−1, collected by the LHCb experiment in 2018 at a nucleon-nucleon centre-of-mass energy s NN = 5.02 TeV. In the mass range from 400 to 1200 MeV, the results are consistent with previous experiments, with the spectrum dominated by the ρ0 meson, which interferes with a nonresonant component, together with a smaller ω meson contribution. In an extended mass range up to 2300 MeV, models previously used do not fit the data and a consistent description requires the introduction of two resonances at masses of 1350 ± 20 MeV and 1790 ± 20 MeV with widths of about 300 MeV. The cross-section for each meson is measured differentially in twelve bins of rapidity from 2.05 to 4.90. The ρ0 cross-section increases with rapidity from about 400 to 600 mb and is measured with a typical precision of 8%, while the cross-section times branching fraction for the ω, ρ′ and ρ′′, with the statistical precision of the data, do not have a pronounced rapidity dependence and are between 0.5 and 1.5mb, with uncertainties up to 30%. A large nuclear suppression is observed for the ρ0 meson compared to expectations based on photoproduction on the proton that use the impulse approximation. Significant suppression is also observed compared to that predicted by elastic scattering described in the Glauber approach, or with the addition of inelastic scattering in a Gribov-Glauber model
Experimental Quantification of the Phonon Drag Deformation Mechanism in Metals at Extreme Strain Rates
No full textExtreme strain rate deformations, above 10⁶ s⁻¹, are seen across many fields of science and engineering; from meteorite impacts and impact induced crystallographic phase changes to high-speed machining and additive manufacturing. Despite the range of applications, many common high-rate impact experiments are intrinsically limited to strain rates of only 10⁴ s⁻¹ before complicating the material deformation with a superimposing state of shock due to high impact pressures. However, recent advances in optically driven microballistics using laser induced projectile impact tests have provided a new quantitative look into extreme mechanics of materials, at rates above 106 s-1 and well below the onset of shock effects.
As deformation strain rates increase, additional strengthening mechanisms in metals become available, leading to a change in the underlying physics of dislocation motion and an increase in strength. This thesis first explores the mechanical properties of pure metals when deformed at extreme strain rates − both in ambient conditions and elevated temperatures. Using an array of complimentary characterization methods, two independent measurements of strength, the dynamic strength and dynamic hardness, are assessed. As the temperature is increased from ambient, the strength and hardness of pure metals both increase an appreciable amount. At these deformation rates, conventional thermal softening effects are now in competition with anti-thermal hardening that arises from ballistic transport of dislocations from phonon interactions in the crystal lattice. These effects are quantified systematically and it is shown that the anomalous thermal strengthening seen is, thermodynamically and kinetically, the expected form of plasticity under these impact conditions.
Next, the limits of where this anomalous thermal strengthening occur in metals are investigated. First, solute elements are added to pure Ni to evaluate how additional dislocation pinning mechanisms effect the strength at ambient and elevated temperatures during extreme strain rate deformations. The strengthen increase due to solute pinning of dislocations is additive to the other strengthening mechanisms, yet thermally controlled, which provides a transition from ballistic transport of dislocations to thermally activated strengthening at a critical concentration of solutes. Finally, the upper bound of temperature for dislocation phonon drag strengthening is assessed. While it was shown that pure metals increase strength with increasing temperature, this “hotter-is-stronger” trend breaks down as the temperature approaches the melting point of the metal. Using Sn, due to its low melting temperature, the breakdown from “hotter-is-stronger” to “hotter-is-softer” as the initial substrate temperature approaches the melting temperature is systematically explored.Ph.D
Iterating Sine, Equivalence Classes of Variable Changes, and Groups with Few Conjugacy Classes
No full textThis is an expository paper about iterations of a
smooth real function f on [0, ) such that f(0) = 0,
f E
(0) = 1, and f(x) 0, i.e., the sequence
defined by xn+1 = f(xn). This sequence has interesting asymptotics, whose study leads to the question of classifying conjugacy classes in the group of formal changes of variable y = f(x), i.e., formal series f(x) = x + a2x2 + a3x2 + ⋯
with real coefficients (under composition). The same classification applies over a finite field p for suitably truncated
series f, defining a family of p-groups that have the smallest
number of conjugacy classes for a given order, i.e., are the
“most noncommutative” finite groups currently known. The
paper should be accessible to undergraduates and at least
partially to advanced high school students
Generative BigSMILES: an extension for polymer informatics, computer simulations & ML/AI
No full textThe BigSMILES notation, a concise tool for polymer ensemble representation, is augmented here by introducing an enhanced version called generative BigSMILES. G-BigSMILES is designed for generative workflows, and is complemented by tailored software tools for ease of use. This extension integrates additional data, including reactivity ratios (or connection probabilities among repeat units), molecular weight distributions, and ensemble size. An algorithm, interpretable as a generative graph is devised that utilizes these data, enabling molecule generation from defined polymer ensembles. Consequently, the G-BigSMILES notation allows for efficient specification of complex molecular ensembles via a streamlined line notation, thereby providing a foundational tool for automated polymeric materials design. In addition, the graph interpretation of the G-BigSMILES notation sets the stage for robust machine learning methods capable of encapsulating intricate polymeric ensembles. The combination of G-BigSMILES with advanced machine learning techniques will facilitate straightforward property determination and in silico polymeric material synthesis automation. This integration has the potential to significantly accelerate materials design processes and advance the field of polymer science
Energy Efficient Real-time Operating Systems on Chip
No full textAutonomous micro-robots are crucial for several tasks, such as search and rescue, noknowledge mapping, and navigation. Without an external power connection, these robots are constrained by their on-platform energy capacity. The power consumption of actuation systems used in micro-robots is within the same magnitude of the power consumption of the compute system. Thus, the remaining factor for enabling these micro-robots is associated with the design of energy-efficient compute systems. Energy usage of compute systems is typically dominated by memory operations, which previous efforts have attempted to mitigate with memory efficient software and hardware. These efforts are enabled with the software/hardware interface, which is implemented as an Operating System (OS). However, Operating Systems for energy-efficient platforms have not been fully explored. Current approaches utilize full general-purpose Operating Systems such as Linux, which can incur large memory and compute overhead penalties. These overheads not only consume the typically limited memory resources of energy-efficient systems, but also increase the number of memory accesses and CPU cycles, both of which are significant contributors to energy consumption. To address these concerns, we propose the design of a computational and memory efficient Real-time Operating System (RTOS). Our RTOS is designed to minimize both memory footprint and compute cycle overhead. It achieves this primarily through direct physical memory access, cycle-efficient task scheduling, and minimal runtime services to avoid unnecessary processing. Additionally, the modular RTOS kernel includes only the components required by an application in the final binary, reducing code size and memory usage without compromising functionality. The design enables the utilization of energy-efficient hardware accelerators and software, allowing for execution of robotics workloads with minimal memory and cycle overhead. When comparing robotics algorithms implemented on our proposed RTOS and baseline OSes, our design was able to achieve a 99% reduction in memory footprint. Additionally, it achieved up to a 47% increase in throughput. Thus, our design demonstrates a direct reduction in memory and CPU cycle overhead, which in turn lowers total system memory and energy consumption. The proposed design was demonstrated and verified on a resource constrained system-on-chip on the AMD Virtex Ultrascale+ VCU118 FPGA.M.Eng
Assessing the Structure-Based Turbulence Model Performance for Thermal Striping Applications Using Symmetric Jet Experiments
No full textTurbulent mixing of coolant streams can result in an oscillatory mixing phenomenon called thermal striping. These fluctuations have the potential to lead to anticipated thermal fatigue failures in advanced nuclear reactors. To predict thermal striping, robust and computationally affordable modeling tools that are capable of accurately representing complex turbulence are needed. Hybrid turbulence approaches, such as detached-eddy simulation and scale-adaptive simulation, have shown some success in resolving complex unsteady turbulence for massively separated flows, however the applicability of these models to internal flows is limited. A STRUCTure-based (STRUCT) second-generation Unsteady Reynolds-Averaged Navier–Stokes turbulence model was recently proposed at the Massachusetts Institute of Technology to robustly extend the applicability of hybrid closures. In this work, the STRUCT model is evaluated using experimental data taken at the Reactor Cavity Cooling System separate-effects test facility at the University of Michigan. The experiments observed the interaction of parallel symmetric rectangular jets, and include measurements for mean profiles of velocity and Reynolds stresses. In the present work, the simulation results are assessed against mean profiles of velocity and Reynolds stresses, demonstrating the ability to reproduce the unsteadiness of the jets in close agreement with the measurements at considerably reduced computational cost
Machine Learning Methods for Churn Prediction and Infrastructure Resilience
No full textThis thesis investigates how advanced machine learning methods can effectively address two critical business challenges facing the telecommunications industry: short-term customer churn prediction and long-term infrastructure resilience to climate-driven disruptions.
In the first part of this work, I develop an upgrades-informed churn forecasting model tailored specifically for marketing operations. Recognizing limitations in the existing aggregate forecasting methodologies, I create a cohort-based cascade model that explicitly integrates customer upgrade behavior across various contract tenures. To address data sparsity and longitudinal gaps in newer contract types, I employ synthetic data generation and imputation techniques, such as regression-based methods and Multivariate Imputation by Chained Equations (MICE). For forecasting churn and upgrade rates, I prioritize interpretability by applying linear regression enhanced with time-series forecasting techniques and macroeconomic indicators, including the Consumer Price Index. This approach significantly improves forecasting accuracy, aligns internal stakeholder objectives, and supports strategic decision-making around customer retention and promotional offers.
The second part focuses on building predictive models and strategic frameworks for long-term infrastructure resilience in the face of increasing climate risks. Leveraging spatial-temporal clustering methods (DBSCAN) and advanced neural network architectures, I develop a model to attribute historical outages to extreme weather events. Further, I integrate this model with future climate scenarios from CMIP5 projections using Monte Carlo simulations, providing actionable insights into future infrastructure vulnerabilities. Employing SHapley Additive exPlanations (SHAP), I interpret model predictions, highlighting critical factors such as precipitation, windspeed, and atmospheric pressure. Additionally, I propose frameworks for quantifying financial impacts of future outages and recommend optimization strategies for proactive infrastructure hardening and emergency response.
Collectively, these applications demonstrate the value of strategically employing interpretable and robust machine learning methodologies to enhance short-term operational decisions and long-term strategic planning within telecom organizations.M.B.A.S.M
Advancing the Science of Teaching with Tutoring Data: A Collaborative Workshop with the National Tutoring Observatory
No full textL@S ’25, Palermo, ItalyEffective teaching is among the most powerful influences on student learning, but scientific progress in understanding effective teaching moves has been held back by insufficient data on teaching. Despite extensive research efforts, progress is hindered by persistent challenges related to data de-identification and preprocessing, annotation and segmentation, multimodal analysis, predictive and causal modeling of student outcomes. Addressing these barriers requires a concerted, interdisciplinary approach. The National Tutoring Observatory (NTO) is a first-of-its-kind research infrastructure designed to unite researchers, developers, tutoring providers, and educational organizations in tackling common barriers to uncovering the dynamics of effective tutoring moves. The NTO is spearheading the creation of the Million Tutor Moves dataset, the largest open-access collection of tutoring interactions, leveraging artificial intelligence to unlock insights that accelerate the science of teaching at scale. This workshop aims to bring together the Learning at Scale community to share progress, identify common challenges, and explore collaborative solutions. The agenda will feature presentations of accepted papers, interactive demos, and a moderated panel bringing together researchers, developers, and tutoring providers. This workshop aims to advance a shared vision for uncovering the fundamental principles of impactful tutoring and teaching through the power of collaborative research and data-driven discovery
Theoretical Foundations for Learning in Games and Dynamic Environments
No full textDecision-making problems lie at the heart of numerous aspects of human and algorithmic behavior across our society, ranging from healthcare systems to financial systems to interactions with the physical world. A central challenge that arises across many decision-making problems is the presence of multiple agents, often with competing incentives. To understand how agents will act in such situations, it is often productive to compute equilibria, which have the property that no agent can deviate from them and improve their utility. An additional challenge is that decisions made by agents often change the state of the environment, which is modeled as dynamic. Thus, we need efficient algorithms for learning good policies, which tell the agent what to do as a function of the environment’s state. Extensive work spanning multiple domains such as economics, computer science, and statistics has been developed to model these decision-making problems. This has led to many celebrated results, which include, for instance, a considerable body of work studying the computational properties of Nash equilibria in normal-form games, and a long line of papers on reinforcement learning. However, many of these classical works suffer from a few shortcomings: first, they often do not account for the enormous state or action spaces available to agents in realistic decision-making settings, and second, many of them do not derive computationally efficient algorithms for the desired solution concepts. These shortcomings are brought to the forefront by the remarkable recent progress in artificial intelligence, which holds promise for solving decision-making problems with enormous state or action spaces but which is often bottlenecked by computation. The objective of this thesis is to develop theoretical foundations for the computational aspects of such decision-making problems: e.g., How do we efficiently compute equilibria in large games?, and: How can we efficiently learn near-optimal policies in complex environments? Some highlights of our results are listed below—first, we study problems in which there are multiple agents and the goal is to compute some notion of equilibrium: • We show the first near-optimal rate of convergence to equilibrium for a no-regret learning algorithm in normal-form games, resolving a decade-long line of work which had aimed to establish increasingly better rates. • We establish the first algorithm with sublinear swap regret against arbitrary adversaries enjoying only polylogarithmic dependence on the number of actions, resolving a question of Blum and Mansour from 2007. • As a corollary of the preceding result, we obtain the first polynomial-time algorithm for approximating a correlated equilibrium in extensive-form games (to constant approximation error), addressing a question of von Stengel & Forges from 2008. Additionally we obtain near-optimal bounds on the communication and query complexity of approximating correlated equilibria in normal-form games (to constant approximation error), addressing several open problems in the literature. • We give the first algorithm for the sequential calibration problem with calibration error beating that of the seminal work of Foster & Vohra from 1998. Moving on to decision-making problems where the environment is modeled as dynamic (typically studied in the framework of reinforcement learning (RL)), our results include the following: • We give the first end-to-end computationally efficient algorithms for learning a nearoptimal policy in many fundamental reinforcement learning problems, such as those of (constant-action) Linear Bellman Complete MDPs and sparse linear MDPs. • We give the first quasi-polynomial time algorithm for finding a near-optimal policy in a general and well-motivated class of partially observable RL environments, and show that our bound is tight. • We prove some (perhaps surprising) hardness results that arise in multi-agent RL problems. For instance, we show that it is computationally hard to implement noregret learning algorithms in multi-agent RL environments even when the agents can coordinate on their choice of algorithm, which creates a stark contrast with simpler multi-agent learning settings (e.g., in normal-form games) where no-regret learning has formed the bedrock for a wide array of developments over the last several decades. • Nevertheless, we show that by adjusting the type of equilibrium appropriately, we can circumvent the above hardness results and derive computationally efficient decentralized algorithms for computing equilibria in multi-agent RL environments. Many of the above results have inspired follow-up work which includes applications of our results to various problems in game theory, reinforcement learning, online learning, and related domains, as well as the formulation of new problems which are inspired by the above results.Ph.D
Eliminating Hallucination-Induced Errors in Code Generation with Functional Clustering
No full textModern code–generation LLMs can already solve a large fraction of programming problems, yet they still hallucinate subtle bugs that make their outputs unsafe for autonomous deployment. We present functional clustering, a black-box wrapper that eliminates nearly all hallucination-induced errors while providing a tunable confidence score. The wrapper samples many candidate programs, executes each on a self-generated test suite, and clusters candidates whose I/O behavior is identical; the empirical mass of the largest cluster serves as an exact confidence estimate. A single scalar threshold on this estimate lets users trade coverage for reliability with exponential guarantees. On LiveCodeBench our verifier preserves baseline pass@1 on solvable tasks yet slashes the error rate of returned answers from ∼65% to 2%, and drives it to 0% at a conservative threshold while still answering 15.6% of prompts. Manual audits show that the few residual mistakes stem from prompt misinterpretation, not random generation noise, narrowing future work to specification clarity. Because the method requires only sampling and sandbox execution, it applies unchanged to closed-source APIs and future models, offering a practical path toward dependable, autonomous code generation.M.Eng