SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
PV Performance Loss Rate Estimation using Spatio-temporal Graph Neural Networks PVplr-stGNN is a Python 3 package developed by the SDLE Research Center at Case Western Reserve University in Cleveland OH. This repository contains the full source PVplr-stGNN package. The package contains the PV-stGAE for missingness data detection and imputation and PV-DynGNN for PLR estimation.
This paper develops the geometric sector of the replication-driven cosmogenesis framework introduced in Paper I. Starting from a pre-geometric spectral substrate and a minimal set of replication axioms, we show how coherent self-replicating units generate a spatial adjacency graph whose continuum limit acquires an effective Riemannian structure. The replication dynamics determines a characteristic correlation length that seeds the local metric, while overlap relations among coherent units produce an isotropic neighborhood geometry with an emergent dimensionality $d_{\rm eff}\simeq 3$ across a broad range of replication factors. As replication slows and causal order stabilizes, a limiting signal speed $c_\ast$ appears, providing the basis for the Lorentzian structure of spacetime without assuming a pre-existing light cone. We derive conditions under which the adjacency graph converges to a smooth three-dimensional manifold, describe the transition from Euclidean to Lorentzian propagation, and identify geometric invariants controlled by the replication parameters. This work establishes the geometric and causal layer of the replication cosmogenesis program, bridging the spectral axioms of Paper I to the cosmological dynamics explored in Paper III.
This report investigates the use of unsupervised probabilistic learning techniques for the analysis of hypersonic trajectories. The algorithm first extracts the intrinsic structure in the data via a diffusion map approach. Using the diffusion coordinates on the graph of training samples, the probabilistic framework augments the original data with samples that are statistically consistent with the original set. The augmented samples are then used to construct conditional statistics that are ultimately assembled in a path-planing algorithm. In this framework the controls are determined stage by stage during the flight to adapt to changing mission objectives in real-time. A 3DOF model was employed to generate optimal hypersonic trajectories that comprise the training datasets. The diffusion map algorithm identfied that data resides on manifolds of much lower dimensionality compared to the high-dimensional state space that describes each trajectory. In addition to the path-planing worflow we also propose an algorithm that utilizes the diffusion map coordinates along the manifold to label and possibly remove outlier samples from the training data. This algorithm can be used to both identify edge cases for further analysis as well as to remove them from the training set to create a more robust set of samples to be used for the path-planing process.
CATCH CLI focuses on gathering telemetry data, storing it in the Neo4j database, querying for Mitre ATT&CK patterns, and creating STIX 2.1 reports. Key Components: Analysis Modules: Analyze data to detect attack patterns. GoSTOTS Collection Engines: Collect telemetry data. These tools can be used together or individually. Analysis modules rely on data from specific engines to identify attack patterns. Source Code Organization: Engines: CATCH/catch/cmd/collection Modules: CATCH/catch/cmd/analysis CGUI Overview CATCH Graphical User Interface (CGUI) offers a graphical shell to execute CATCH CLI, allowing easy editing of: Analysis Modules Database configurations Profiles (collection and device settings) Neo4j Overview Neo4j is a graph database using the Cypher query language, storing data in JSON. It seamlessly integrates with STIX 2.1 data for: Data Submission: CATCH Collection Engines Data Querying: Analysis Modules CATCH modifies STIX 2.1 data for Neo4j submission and reverts it back during querying. STIG Overview Structured Threat Intelligence Graph (STIG) is a tool for creating, editing, querying, analyzing, and visualizing threat intelligence using STIX 2.1 and storing data in Neo4j. Usage Tools can be run: Manually (CLI): Refer to CATCH documentation User Interface: Run ./cgui/CGUI or go run ./cgui/ Additional Information Logging System: Detailed in the config documentation Further Documentation: Available for CATCH and CGUI
Machine learning (ML) offers considerable promise for the design of new molecules and materials. In real-world applications, the design problem is often domain-specific, and suffers from insufficient data, particularly labeled data, for ML training. In this study, we report a data-efficient, deep-learning framework for molecular discovery that integrates a coarse-grained functional-group representation with a self-attention mechanism to capture intricate chemical interactions. Our approach exploits group-contribution concepts to create a graph-based intermediate representation of molecules, serving as a low-dimensional embedding that substantially reduces the data demands typically required for training. Using a self-attention mechanism to learn the subtle but highly relevant chemical context of functional groups, the method proposed here consistently outperforms existing approaches for predictions of multiple thermophysical properties. In a case study focused on adhesive polymer monomers, we train on a limited dataset comprising only 6,000 unlabeled and 600 labeled monomers. The resulting chemistry prediction model achieves over 92% accuracy in forecasting properties directly from SMILES strings, exceeding the performance of current state-of-the-art techniques. Furthermore, the latent molecular embedding is invertible, enabling the design pipeline to automatically generate new monomers from the learned chemical subspace. We illustrate this functionality by targeting several properties, including high and low glass transition temperatures (Tg), and demonstrate that our model can identify new candidates with values that surpass those in the training set. The ease with which the proposed framework navigates both chemical diversity and data scarcity offers a promising route to accelerate and broaden the search for functional materials.
Nominally weakly coordinating anions are useful for modulating the solubility and chemical properties of metal complexes, but identification and analysis of the systematics of the interactions of anions with cationic metal complexes has not received the attention it deserves. Here, a chemical informatics approach is demonstrated for identifying and quantitatively analyzing the ways that the bis(trifluoromethylsulfonyl)imide anion (TFSI) can interact with metal-containing species. An open access computer program (PyCIFTer) was developed to facilitate large-scale structural analysis of TFSI-containing species by utilization of experimental atomic coordinate data from single-crystal X-ray diffraction (XRD) studies obtained from the Cambridge Structural Database (CSD). PyCIFTer establishes a three-dimensional vector space from the raw atomic coordinates, generating acyclic, undirected graphs that are used to rapidly analyze the structural properties (bond lengths and angles) of TFSI in individual structures in sequential/batch fashion. The structures are sorted by PyCIFTer into groups based on pre-set and chemically sensible criteria, affording a comprehensive and systematic view of TFSI structural chemistry. This approach avoids tedious one-at-a-time interrogation of structures, a prospect unreasonable in this case, and many others of contemporary chemical relevance; there were over 1500 structures in the CSD containing TFSI as of November 2024. The results demonstrate that TFSI only rarely binds to cations in the solid state, favoring the formation of species in which TFSI is found in cations’ outer coordination spheres. The prospect of applying PyCIFTer to other moieties is also discussed. PyCIFTer is also schematically compared to the commercial CSD Python application programming interface (API). Taken together, this work demonstrates the usefulness of modular workflows for sequential/batch analysis of structural data from XRD, an approach that appears poised to accelerate the translation of legacy structural results into new chemical insights and hypotheses.
We explore a deep connection between fracton order and product codes. In particular, we propose and analyze conditions on classical seed codes which lead to fracton order in the resulting quantum product codes. Depending on the properties of the input codes, product codes can realize either Type-I or Type-II fracton models, in both nonlocal and local constructions. For the nonlocal case, we show that a recently proposed model of lineons on nonlocal graphs can be obtained as a hypergraph product code. Interestingly, constrained mobility in this model arises only from energy barriers associated with the graph. For the local case, we introduce a novel type of classical LDPC code defined on a planar aperiodic tiling. By considering the specific example of the pinwheel tiling, we demonstrate the systematic construction of local Type-I and Type-II fracton models as product codes. Our work establishes product codes as a natural setting for exploring fracton order.
In a large hospital system, a network of hospitals relies on electronic health records (EHRs) to make informed decisions regarding their patients in various clinical domains. Consequently, the dependability of the health information technology (HIT) systems responsible for collecting EHR data is of utmost importance for patient safety. Recently, novel methods and tools aimed at identifying anomalies in EHR data to bolster the reliability of HIT systems have been introduced. However, these existing methods and tools primarily concentrate on individual hospitals, which limits our understanding of system-wide anomalous events and their potential impact on patient safety across multiple hospitals. In this article, we introduce a new approach to detecting anomalies in EHR data within a network of hospitals. This is achieved by combining advanced machine learning techniques with graph algorithms to create a tool capable of swiftly identifying and responding to deviations. Our proposed approach employs a combination of five machine learning models, harnessing the unique strengths of each model to provide a more robust detection system. The detected anomalies are then represented as graphs, allowing us to recognize patterns across the hospital network. This aids in identifying anomalies that span multiple medical facilities, potentially indicating broader system-level risks. Extensive real-world testing of our approach demonstrated its ability to offer actionable insights compared to existing methods. Additionally, its scalable design ensures seamless integration into existing HIT infrastructures.
Microbiome data require statistical models that can simultaneously decode microbes' reaction to the environment and interactions among microbes. While a multiresponse linear regression model seems like a straight-forward solution, we argue that treating it as a graphical model is problematic given that the regression coefficient matrix does not encode the conditional dependence structure between response and predictor nodes. This observation is especially important in biological settings when we have prior knowledge on the edges from specific experimental interventions that can only be properly encoded under a conditional dependence model. Here, we propose a chain graph model with two sets of nodes (predictors and responses) whose solution yields a graph with edges that indeed represent conditional dependence, thus agreeing with the experimenter's intuition on the average behavior of nodes under treatment. The solution to our model is sparse via the Bayesian linear regression (LASSO). In addition, we propose an adaptive extension so that different shrinkages can be applied to different edges to incorporate edge-specific prior knowledge. Our model is computationally inexpensive through an efficient Gibbs sampling algorithm and can account for binary, counting, and compositional responses via an appropriate hierarchical structure. We test the performance of our model in a variety of simulated datasets, thereby showing superior performance to state-of-the-art approaches. We further apply our model to human gut and soil microbial compositional datasets, and we highlight that CG-LASSO can estimate biologically meaningful network structures in the data.
The ability of machine learning (ML) classification models to resist small, targeted input perturbations—known as adversarial attacks—is a key measure of their safety and reliability. We show that floating-point non associativity (FPNA) coupled with asynchronous parallel programming on GPUs is sufficient to result in misclassification, without any perturbation to the input. Additionally, we show that this misclassification is particularly significant for inputs close to the decision boundary and that standard adversarial robustness results may be overestimated up to 4.6 when not considering machine-level details. We first study a linear classifier, before focusing on standard Graph Neural Network (GNN) architectures and datasets used in robustness assessments. We develop a novel black-box attack using Bayesian optimization to discover external workloads that can change the instruction scheduling which bias the output of reductions on GPUs and reliably lead to misclassification. Motivated by these results, we present a new learnable permutation (LP) gradient-based approach to learning floating-point operation orderings that lead to misclassifications. The LP approach provides a worst-case estimate in a computationally efficient manner, avoiding the need to run identical experiments tens of thousands of times over a potentially large set of possible GPU states or architectures. Finally, using instrumentation-based testing, we investigate parallel reduction ordering across different GPU architectures under external background workloads, when utilizing multi-GPU virtualization, and when applying power capping. Our results demonstrate that parallel reduction ordering varies significantly across architectures under the first two conditions, substantially increasing the search space required to fully test the effects of this parallel scheduler-based vulnerability. These results and the methods developed here can help to include machine-level considerations into adversarial robustness assessments, which can make a difference in safety and mission critical applications.
Benders decomposition is a widely used method for solving large and structured optimization problems, but its performance is affected by the repeated solution of subproblems. We propose a flexible and modular algorithmic framework for accelerating Benders decomposition. Specifically, we express the problem structure by using a graph-theoretic modeling abstraction in which nodes represent optimization subproblems and edges represent connectivity between subproblems. A key innovation of our approach is that we embed multiparametric programming (mp) surrogates for node subproblems, which maps the exact analytical map of the subproblem solution space. The use of mp surrogates allows us to replace subproblem solves with fast look-ups and function evaluations for primal and dual variables during the iterative Benders process. We formally show the equivalence between classical Benders cuts and those derived from the mp solution. We implement our framework in the open-source PlasmoBenders.jl software package. To demonstrate the capabilities of the proposed framework, we apply it to a two-stage stochastic programming problem, which aims to make optimal capacity expansion decisions under market uncertainty. We evaluate both single-cut and multicut variants of Benders decomposition and show that the use of mp surrogates achieves substantial speedups in subproblem solve time, while preserving the convergence guarantees of Benders decomposition. We highlight advantages in solution analysis and interpretability that is enabled by mp critical region tracking; specifically, we show that these reveal how decisions evolve geometrically across the Benders search. Our results aim to demonstrate that combining surrogate modeling with graph modeling offers a promising and extensible foundation for structure-exploiting decomposition. In addition, by decomposing the problem into more tractable subproblems, the proposed approach also aims to overcome scalability issues of mp. Finally, the use of mp surrogates provides a unifying and modular optimization framework that enables the representation of heterogeneous node subproblems as modeling objects with a homogeneous structure.
Remote detectability is often taken as a physical assumption in the study of topologically ordered systems, and it is a central axiom of mathematical frameworks of topological quantum field theories. We show under the entanglement bootstrap approach that remote detectability is a necessary property; that is, we derive it as a theorem. Starting from a single wave function on a topologically-trivial region satisfying the entanglement bootstrap axioms, we can construct states on closed manifolds. The crucial technique is to immerse the punctured manifold into the topologically trivial region and then heal the puncture. This is analogous to Kirby’s torus trick. We then analyze a special class of such manifolds, which we call pairing manifolds. For each pairing manifold, which pairs two classes of excitations, we identify an analog of the topological S S -matrix. This pairing matrix is unitary, which implies remote detectability between two classes of excitations. These matrices are in general not associated with the mapping class group of the manifold. As a by-product, we can count excitation types (e.g., graph excitations in 3+1d). The pairing phenomenon occurs in many physical contexts, including systems in different dimensions, with or without gapped boundaries. We provide a variety of examples to illustrate its scope.
The critical infrastructures protection landscape is a vast and varied pattern of independent, but interconnected infrastructure systems that are essential to the function of our modern society. The U.S. policy on critical infrastructure protection has been continually evolving since the “President’s Commission on Critical Infrastructure Protection” was published in 1997. In response to these policies, federal, state, and local governments, along with research institutions, have invested a substantial amount of time and effort into identifying and analyzing critical infrastructure, their functions, and dependencies/interdependencies to better understand their vulnerabilities. To date, the ability to assess vulnerabilities, resiliency, and priorities for protecting interdependent critical infrastructure systems from an all-hazards perspective remains a difficult problem. In this paper we introduce the All-Hazards Analysis (AHA) methodology, which provides an integrated functional basis across infrastructure systems, through the implementation of a common language and a scalable level of decomposition to effectively evaluate the resilience of interconnected infrastructure systems. AHA models infrastructure systems as directed multidimensional graphs, which enable the evaluation of cross-sector interdependencies prior to, during, and after disruptive events. Finally, and by design, AHA enables the cross linking of data taxonomies to enable more effective data sharing, such as the National Critical Functions (NCF) and Infrastructure Data Taxonomy (IDT).
Abstract Single-cell RNA sequencing (scRNA-seq) permits researchers to study the complex mechanisms of cell heterogeneity and diversity. Unsupervised clustering is of central importance for the analysis of the scRNA-seq data, as it can be used to identify putative cell types. However, due to noise impacts, high dimensionality and pervasive dropout events, clustering analysis of scRNA-seq data remains a computational challenge. Here, we propose a new deep structural clustering method for scRNA-seq data, named scDSC, which integrate the structural information into deep clustering of single cells. The proposed scDSC consists of a Zero-Inflated Negative Binomial (ZINB) model-based autoencoder, a graph neural network (GNN) module and a mutual-supervised module. To learn the data representation from the sparse and zero-inflated scRNA-seq data, we add a ZINB model to the basic autoencoder. The GNN module is introduced to capture the structural information among cells. By joining the ZINB-based autoencoder with the GNN module, the model transfers the data representation learned by autoencoder to the corresponding GNN layer. Furthermore, we adopt a mutual supervised strategy to unify these two different deep neural architectures and to guide the clustering task. Extensive experimental results on six real scRNA-seq datasets demonstrate that scDSC outperforms state-of-the-art methods in terms of clustering accuracy and scalability. Our method scDSC is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DHUDBlab/scDSC.
Significant progress has been made in our understanding of the analytic structure of FRW wavefunction coefficients, facilitated by the development of efficient algorithms to derive the differential equations they satisfy. Moreover, recent findings indicate that the twisted cohomology of the associated hyperplane arrangement defining FRW integrals overestimates the number of integrals required to define differential equations for the wave-function coefficient. We demonstrate that the associated dual cohomology is automatically organized in a way that is ideal for understanding and exploiting the cut/residue structure of FRW integrals. Utilizing this understanding, we develop a systematic approach to organize compatible sequential residues, which dictates the physical subspace of FRW integrals for any n -site, ℓ-loop graph. In particular, the physical subspace of tree-level FRW wavefunction coefficients is populated by differential forms associated to cuts/residues that factorize the integrand of the wavefunction coefficient into only flat space amplitudes. After demonstrating the validity of our construction using intersection theory, we develop simple graphical rules for cut tubings that enumerate the space of physical cuts and, consequently, differential forms without any calculation.
Graph neural networks (GNNs), an emerging class of machine learning models for graphs, have gained popularity for their superior performance in various graph analytical tasks. Mini-batch training is commonly used to train GNNs on large graphs, and data parallelism is the standard approach to scale mini-batch training across multiple GPUs. Data parallel approaches contain redundant work as subgraphs sampled by different GPUs contain significant overlap. To address this issue, we introduce a hybrid parallel mini-batch training paradigm called Split parallelism. Split parallelism avoids redundant work by splitting the sampling, loading, and training of each mini-batch across multiple GPUs. Split parallelism, however, introduces communication overheads that can be more than the savings from removing redundant work. We further present a lightweight partitioning algorithm that probabilistically minimizes these overheads. We implement spllit parllelism in GSplit and show that it outperforms state-of-the-art mini-batch training systems like DGL, Quiver, and P3.
Neutrinos are the second most abundant massive particle in the universe, but because of their small mass and weak interactions with matter, they are still not well understood. Due to their pervasive nature and unique properties, neutrinos make great probes to explore fundamental questions about the universe such as what is the composition of dark matter? The MiniBooNE experiment, designed to study neutrino oscillations, detected what has become a longstanding anomaly – an excess of low energy electromagnetic events from accelerator neutrino interactions. A possible explanation of the anomaly comes from neutrino interactions with nuclei that create heavier, sterile neutrinos that will decay quickly into electron-positron pairs. Complementing this is the MicroBooNE experiment, a liquid argon time projection chamber, that can reconstruct events well in both two and three dimensions and can distinguish electrons from photons. This reconstruction ability enables us to select dark sector neutrino interactions from other possible interactions using machine learning techniques such as graph neural networks (GNNs) which is the primary focus of this project.
Thin film deposition is a fundamental technology for the discovery, optimization, and manufacturing of functional materials. Deposition by molecular beam epitaxy (MBE) typically employs reflection high-energy electron diffraction (RHEED) as a real-time in situ probe of the growing film. However, the state-of-the-art for RHEED analysis during deposition requires human observation. Here, we present an approach using machine learning (ML) methods to monitor, analyze, and interpret RHEED images on-the-fly during thin film deposition. In the analysis workflow, RHEED pattern images are collected at one frame per second and featurized using a pretrained deep convolutional neural network. The feature vectors are then statistically analyzed to identify changepoints; these changepoints can be related to changes in the deposition mode from initial film nucleation to a transition regime, smooth film deposition, and in some cases, an additional transition to a rough, islanded deposition regime. The feature vectors are additionally analyzed via graph analysis and community classification. The graph is quantified as a stabilization plot, and we show that inflection points in the stabilization plot correspond to changes in the growth regime. The full RHEED analysis workflow is termed RHAAPsody and includes data transfer and output to a visual dashboard. We demonstrate the functionality of RHAAPsody by analyzing the precaptured RHEED images from epitaxial depositions of anatase TiO2 on SrTiO3(001) and show that the analysis workflow can be executed in less than 1 s. Our approach shows promise as one component of ML-enabled real-time feedback control of the MBE deposition process.