arXiv subjects

Matthew West

Publications and source records attributed to Matthew West.

At least 19 recordsLinked to original sources

arXiv2026

Consistently Good vs. Occasionally Great: A Rubric for Open-Ended Feedback Quality from Humans and Machines

Providing high-quality feedback on student work is essential for learning, yet delivering such feedback at scale remains challenging. In this paper, we focus on feedback for open-ended short answer questions in introductory programming, with the goal of nudging students toward success on reattempts without revealing the correct answer. We develop a five-criteria rubric grounded in educational literature for evaluating feedback quality: (1) acknowledging correct portions of the student answer, (2) identifying at least one flaw (if present), (3) providing actionable guidance for improvement, (4) maintaining appropriate concealment of the answer, and (5) using an appropriate conversational tone. Using this rubric, we compare feedback generated by a frontier LLM (OpenAI o1) to feedback from nine teaching assistants across 90 student responses, with three researchers and an LLM independently scoring all feedback. Our results show that while one TA often produced the best feedback, the LLM demonstrated consistently higher average performance than TAs, as evaluated by humans. However, we also uncover significant self-preference bias when using LLMs to evaluate feedback quality: the LLM systematically rated its own outputs higher than human experts did. This bias, which research suggests persists even in cross-model evaluation, raises important methodological concerns for researchers employing LLM-based evaluation. We provide detailed characterization of both TA and LLM performance, analyze sources of variance in TA feedback quality, and discuss implications for deploying LLM-generated feedback in educational settings.

cs.CY↗

arXiv2026

AeroMELD: A Linear Embedding of Aerosol Populations for Diagnostics and Latent Dynamics

Accurately representing atmospheric aerosol populations is essential for simulating aerosol-cloud interactions, radiative forcing, and ice nucleation, yet existing reduced schemes impose structural assumptions that limit their ability to capture composition diversity and mixing state. Machine-learning approaches offer more flexible representations, but standard autoencoders do not preserve the mathematical structure of aerosol populations and therefore cannot support physically meaningful process operators. We introduce AeroMELD (Aerosol Measure Embedding for Latent Dynamics), a mathematically grounded framework for constructing low-dimensional latent variables that retain this structure. We show that any permutation-invariant linear encoder must take a scale-shape decomposition, with total number concentration represented explicitly and latent shape given by a barycentric combination of per-particle embeddings. This aggregated latent state retains the diagnostic expressiveness of a Deep Sets model by moving the nonlinear post-aggregation stage into the learned diagnostic map while preserving latent linearity. Using particle-resolved data as ground truth, we encode weighted particle populations directly rather than binned aerosol states; size-resolved mass and number distributions serve only as diagnostic targets and visual summaries. The latent space accurately reconstructs these distributions, CCN spectra, optical coefficients, and immersion-freezing behavior while preserving the linear population structure needed for hybrid ML-physics models. Although the experiments focus on diagnostic reconstruction, the embedding is designed so that emissions and mixing can be represented exactly and nonlinear microphysical processes learned in a controlled latent space. This work establishes a foundation for learning aerosol-process evolution directly in latent space.

cs.LG↗

arXiv2026

Automated Grading of Handwritten Mathematics Using Vision-Capable LLMs

Automated grading systems have enabled scalable assessment for many response types, but handwritten mathematics remains a barrier due to the complexity of multi-step solutions. Vision-capable large language models (LLMs) offer new opportunities here, yet their reliability in authentic instructional settings remains poorly understood. We present an empirical evaluation of an LLM-based grader for handwritten mathematical work using instructor-defined rubrics. Extending a prior pipeline for typed responses, we integrate transcription and rubric-based evaluation of photographic submissions within a single LLM call, evaluating on student work from two university STEM courses. Comparing AI grading decisions against human-assigned ground truth at the rubric-item level, we observe high overall accuracy, with most errors -- 87\% in the best model -- attributable to transcription failures rather than rubric misapplication. We categorize common error modes, including image quality issues, hallucinated content, and incorrect handling of equivalent expressions. These findings highlight both the promise and limitations of LLM-based grading for handwritten mathematics, providing guidance for system design, prompt refinement, and deployment in educational settings.

cs.CY↗

arXiv2025

Generative Modeling of Aerosol State Representations

Aerosol-cloud--radiation interactions remain among the most uncertain components of the Earth's climate system, in partdue to the high dimensionality of aerosol state representations and the difficulty of obtaining complete \textit{in situ} measurements. Addressing these challenges requires methods that distill complex aerosol properties into compact yet physically meaningful forms. Generative autoencoder models provide such a pathway. We present a framework for learning deep variational autoencoder (VAE) models of speciated mass and number concentration distributions, which capture detailed aerosol size-composition characteristics. By compressing hundreds of original dimensions into ten latent variables, the approach enables efficient storage and processing while preserving the fidelity of key diagnostics, including cloud condensation nuclei (CCN) spectra, optical scattering and absorption coefficients, and ice nucleation properties. Results show that CCN spectra are easiest to reconstruct accurately, optical properties are moderately difficult, and ice nucleation properties are the most challenging. To improve performance, we introduce a preprocessing optimization strategy that avoids repeated retraining and yields latent representations resilient to high-magnitude Gaussian noise, boosting accuracy for CCN spectra, optical coefficients, and frozen fraction spectra. Finally, we propose a novel realism metric -- based on the sliced Wasserstein distance between generated samples and a held-out test set -- for optimizing the KL divergence weight in VAEs. Together, these contributions enable compact, robust, and physically meaningful representations of aerosol states for large-scale climate applications.

physics.ao-ph↗

arXiv2025

Unstructured to structured: geometric multigrid on complex geometries via domain remapping

For domains that are easily represented by structured meshes, robust geometric multigrid solvers can quickly provide the numerical solution to many discretized elliptic PDEs. However, for complicated domains with unstructured meshes, constructing suitable hierarchies of meshes becomes challenging. We propose a framework for mapping computations from such complex domains to regular computational domains via diffeomorphisms, enabling the use of robust geometric-style multigrid. This mapping facilitates regular memory accesses during solves, improving efficiency and scalability, especially on massively parallel processors such as GPUs. Moreover, we show that the diffeomorphic mapping itself may be approximately learned using an invertible neural network, facilitating automated application to geometries where no analytic mapping is readily available.

math.NA↗

arXiv2025

Teaching An Old Dog New Tricks: Porting Legacy Code to Heterogeneous Compute Architectures With Automated Code Translation

Legacy codes are in ubiquitous use in scientific simulations; they are well-tested and there is significant time investment in their use. However, one challenge is the adoption of new, sometimes incompatible computing paradigms, such as GPU hardware. In this paper, we explore using automated code translation to enable execution of legacy multigrid solver code on GPUs without significant time investment and while avoiding intrusive changes to the codebase. We developed a thin, reusable translation layer that parses Fortran 2003 at compile time, interfacing with the existing library Loopy to transpile to C++/GPU code, which is then managed by a custom MPI runtime system that we created. With this low-effort approach, we are able to achieve a payoff of an approximately 2-3x speedup over a full CPU socket, and 6x in multi-node settings.

cs.DC↗

arXiv2024

Leveraging Pre-trained and Transformer-derived Embeddings from EHRs to Characterize Heterogeneity Across Alzheimer's Disease and Related Dementias

Alzheimer's disease is a progressive, debilitating neurodegenerative disease that affects 50 million people globally. Despite this substantial health burden, available treatments for the disease are limited and its fundamental causes remain poorly understood. Previous work has suggested the existence of clinically-meaningful sub-types, which it is suggested may correspond to distinct etiologies, disease courses, and ultimately appropriate treatments. Here, we use unsupervised learning techniques on electronic health records (EHRs) from a cohort of memory disorder patients to characterise heterogeneity in this disease population. Pre-trained embeddings for medical codes as well as transformer-derived Clinical BERT embeddings of free text are used to encode patient EHRs. We identify the existence of sub-populations on the basis of comorbidities and shared textual features, and discuss their clinical significance.

cs.LG↗

arXiv2023

Measuring the Impact of Distractors on Student Learning Gains while Using Proof Blocks

Background: Proof Blocks is a software tool that enables students to construct proofs by assembling prewritten lines and gives them automated feedback. Prior work on learning gains from Proof Blocks has focused on comparing learning gains from Proof Blocks against other learning activities such as writing proofs or reading. Purpose: The study described in this paper aims to compare learning gains from different variations of Proof Blocks. Specifically, we attempt to quantify the difference in learning gains for students who complete Proof Blocks problems with and without distractors. Methods: We conducted a randomized controlled trial with three experimental groups: a control group that completed an off-topic Proof Blocks activity, one that completed a \tool{} activity without distractors, and one that completed a Proof Blocks activity with distractors. All three groups read a book chapter on proof by induction before completing their activity. Findings: The group that completed the Proof Blocks activity with distractors performed better on the posttest than the group that completed the Proof Blocks without distractors, who in turn performed better than the group that completed the off-topic Proof Blocks activity. However, none of these differences were statistically significant. While the results of this study are inconclusive, we hope that it can serve as a foundation for future work.

cs.HC↗

arXiv2023

Immersion freezing in particle-based aerosol-cloud microphysics: a probabilistic perspective on singular and time-dependent models

Cloud droplets containing ice-nucleating particles (INPs) may freeze at temperatures above the homogeneous freezing threshold temperature. This process, referred to as immersion freezing, is one of the modulators of aerosol-cloud interactions in the Earth's atmosphere. In modeling studies, immersion freezing is often described using either so-called "singular" or "time-dependent" parameterizations. Here, we juxtapose both approaches and discuss them in the context of probabilistic particle-based cloud microphysics modeling. First, using a box model, we contrast how both parameterizations respond to different idealized ambient cooling rate profiles and quantify the impact of the polydispersity of the immersed surface spectrum on the frozen fraction evolution. Second, using a prescribed-flow two-dimensional cloud model, we illustrate the implications of applying the singular model in simulations with flow regimes relevant to ambient cloud conditions rather than to the cloud-chamber experiments on which these parameterizations are built upon. We discuss the critical role of the attribute-space sampling strategy for particle-based model simulations in modeling heterogeneous ice nucleation which is contingent on the presence of relatively sparse immersed INPs. The key takeaways include: (i) The singular approach, constituting a time-integrated form of a more general time-dependent approach, is only applicable under a limited range of ambient cooling rates. (ii) The time-dependent approach, especially when based on water-activity, is suitable for integration with particle-based model components of detailed aerosol composition and collisional growth/breakup. (iii) A flow-coupled aerosol-budget-resolving simulation shows the benefits and challenges of modeling cloud condensation nuclei activation and immersion freezing on insoluble ice nuclei with super-particle methods.

physics.flu-dyn↗

arXiv2023

PyPartMC: A Pythonic interface to a particle-resolved, Monte Carlo aerosol simulation framework

PyPartMC is a Pythonic interface to PartMC, a stochastic, particle-resolved aerosol model implemented in Fortran. Both PyPartMC and PartMC are free, libre, and open-source. PyPartMC reduces the number of steps and mitigates the effort necessary to install and utilize the resources of PartMC. Without PyPartMC, setting up PartMC requires: working with UNIX shell, providing Fortran and C libraries, and performing standard Fortran and C source code configuration, compilation and linking. This can be challenging for those less experienced with computational research or those intending to use PartMC in environments where provision of UNIX tools is less straightforward (e.g., on Windows). PyPartMC offers a single-step installation/upgrade process of PartMC and all dependencies through the pip Python package manager on Linux, macOS, and Windows. This allows streamlined access to the unmodified and versioned Fortran internals of the PartMC codebase from both Python and other interoperable environments (e.g., Julia through PyCall). Consequently, users of PyPartMC can setup, run, process and visualize output of PartMC simulations using a single general-purpose programming language.

cs.MS↗

arXiv2023

Learning from Integral Losses in Physics Informed Neural Networks

This work proposes a solution for the problem of training physics-informed networks under partial integro-differential equations. These equations require an infinite or a large number of neural evaluations to construct a single residual for training. As a result, accurate evaluation may be impractical, and we show that naive approximations at replacing these integrals with unbiased estimates lead to biased loss functions and solutions. To overcome this bias, we investigate three types of potential solutions: the deterministic sampling approaches, the double-sampling trick, and the delayed target method. We consider three classes of PDEs for benchmarking; one defining Poisson problems with singular charges and weak solutions of up to 10 dimensions, another involving weak solutions on electro-magnetic fields and a Maxwell equation, and a third one defining a Smoluchowski coagulation problem. Our numerical results confirm the existence of the aforementioned bias in practice and also show that our proposed delayed target approach can lead to accurate solutions with comparable quality to ones estimated with a large sample size integral. Our implementation is open-source and available at https://github.com/ehsansaleh/btspinn.

cs.LG↗

arXiv2023

Hierarchical Graph Neural Network with Cross-Attention for Cross-Device User Matching

Cross-device user matching is a critical problem in numerous domains, including advertising, recommender systems, and cybersecurity. It involves identifying and linking different devices belonging to the same person, utilizing sequence logs. Previous data mining techniques have struggled to address the long-range dependencies and higher-order connections between the logs. Recently, researchers have modeled this problem as a graph problem and proposed a two-tier graph contextual embedding (TGCE) neural network architecture, which outperforms previous methods. In this paper, we propose a novel hierarchical graph neural network architecture (HGNN), which has a more computationally efficient second level design than TGCE. Furthermore, we introduce a cross-attention (Cross-Att) mechanism in our model, which improves performance by 5% compared to the state-of-the-art TGCE method.

cs.LG↗

arXiv2023

Generalizing Lloyd's algorithm for graph clustering

Clustering is a commonplace problem in many areas of data science, with applications in biology and bioinformatics, understanding chemical structure, image segmentation, building recommender systems, and many more fields. While there are many different clustering variants (based on given distance or graph structure, probability distributions, or data density), we consider here the problem of clustering nodes in a graph, motivated by the problem of aggregating discrete degrees of freedom in multigrid and domain decomposition methods for solving sparse linear systems. Specifically, we consider the challenge of forming balanced clusters in the graph of a sparse matrix for use in algebraic multigrid, although the algorithm has general applicability. Based on an extension of the Bellman-Ford algorithm, we generalize Lloyd's algorithm for partitioning subsets of Rn to balance the number of nodes in each cluster; this is accompanied by a rebalancing algorithm that reduces the overall energy in the system. The algorithm provides control over the number of clusters and leads to "well centered" partitions of the graph. Theoretical results are provided to establish linear complexity and numerical results in the context of algebraic multigrid highlight the benefits of improved clustering.

math.NA↗

arXiv2023

A Pair of Non-Isometric Potentials With the Same Semiclassical Invariants

We show that there exist pairs of non-isometric potentials for the 1D semiclassical Schr\"odinger operator whose spectra agree up to $O(h^\infty)$, yet their corresponding eigenvalues differ no less than exponentially. This result was conjectured by Guillemen and Hezari in [GH12], where they prove a very similar result, yet cannot remove the possibility of a subsequence $h_k\to 0$ where the ground state eigenvalues may agree.

math-ph↗

arXiv2023

MG-GNN: Multigrid Graph Neural Networks for Learning Multilevel Domain Decomposition Methods

Domain decomposition methods (DDMs) are popular solvers for discretized systems of partial differential equations (PDEs), with one-level and multilevel variants. These solvers rely on several algorithmic and mathematical parameters, prescribing overlap, subdomain boundary conditions, and other properties of the DDM. While some work has been done on optimizing these parameters, it has mostly focused on the one-level setting or special cases such as structured-grid discretizations with regular subdomain construction. In this paper, we propose multigrid graph neural networks (MG-GNN), a novel GNN architecture for learning optimized parameters in two-level DDMs\@. We train MG-GNN using a new unsupervised loss function, enabling effective training on small problems that yields robust performance on unstructured grids that are orders of magnitude larger than those in the training set. We show that MG-GNN outperforms popular hierarchical graph network architectures for this optimization and that our proposed loss function is critical to achieving this improved performance.

cs.LG↗

arXiv2022

Generalizing Reduction-Based Algebraic Multigrid

Algebraic Multigrid (AMG) methods are often robust and effective solvers for solving the large and sparse linear systems that arise from discretized PDEs and other problems, relying on heuristic graph algorithms to achieve their performance. Reduction-based AMG (AMGr) algorithms attempt to formalize these heuristics by providing two-level convergence bounds that depend concretely on properties of the partitioning of the given matrix into its fine- and coarse-grid degrees of freedom. MacLachlan and Saad (SISC 2007) proved that the AMGr method yields provably robust two-level convergence for symmetric and positive-definite matrices that are diagonally dominant, with a convergence factor bounded as a function of a coarsening parameter. However, when applying AMGr algorithms to matrices that are not diagonally dominant, not only do the convergence factor bounds not hold, but measured performance is notably degraded. Here, we present modifications to the classical AMGr algorithm that improve its performance on matrices that are not diagonally dominant, making use of strength of connection, sparse approximate inverse (SPAI) techniques, and interpolation truncation and rescaling, to improve robustness while maintaining control of the algorithmic costs. We present numerical results demonstrating the robustness of this approach for both classical isotropic diffusion problems and for non-diagonally dominant systems coming from anisotropic diffusion.

math.NA↗

arXiv2022

Efficiency of Learning from Proof Blocks Versus Writing Proofs

Proof Blocks is a software tool that provides students with a scaffolded proof-writing experience, allowing them to drag and drop prewritten proof lines into the correct order instead of starting from scratch. In this paper we describe a randomized controlled trial designed to measure the learning gains of using Proof Blocks for students learning proof by induction. The study participants were 332 students recruited after completing the first month of their discrete mathematics course. Students in the study took a pretest and read lecture notes on proof by induction, completed a brief (less than 1 hour) learning activity, and then returned one week later to complete the posttest. Depending on the experimental condition that each student was assigned to, they either completed only Proof Blocks problems, completed some Proof Blocks problems and some written proofs, or completed only written proofs for their learning activity. We find that students in the early phases of learning about proof by induction are able to learn just as much from reading lecture notes and using Proof Blocks as by reading lecture notes and writing proofs from scratch, but in far less time on task. This finding complements previous findings that Proof Blocks are useful exam questions and are viewed positively by students.

cs.CY↗

arXiv2022

How does a Rational Agent Act in an Epidemic?

Evolution of disease in a large population is a function of the top-down policy measures from a centralized planner, as well as the self-interested decisions (to be socially active) of individual agents in a large heterogeneous population. This paper is concerned with understanding the latter based on a mean-field type optimal control model. Specifically, the model is used to investigate the role of partial information on an agent's decision-making, and study the impact of such decisions by a large number of agents on the spread of the virus in the population. The motivation comes from the presymptomatic and asymptomatic spread of the COVID-19 virus where an agent unwittingly spreads the virus. We show that even in a setting with fully rational agents, limited information on the viral state can result in an epidemic growth.

math.OC↗