A multi-modal pre-training transformer for universal transfer learning in metal–organic frameworks
Not provided.
Computer Science↗
SEARCH · Engineering Papers
Results for “Computation science”
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.
Topological structure of complex predictions
Abstract Current complex prediction models are the result of fitting deep neural networks, graph convolutional networks or transducers to a set of training data. A key challenge with these models is that they are highly parameterized, which makes describing and interpreting the prediction strategies difficult. We use topological data analysis to transform these complex prediction models into a simplified topological view of the prediction landscape. The result is a map of the predictions that enables inspection of the model results with more specificity than dimensionality-reduction methods such as tSNE and UMAP. The methods scale up to large datasets across different domains. We present a case study of a transformer-based model previously designed to predict expression levels of a piece of DNA in thousands of genomic tracks. When the model is used to study mutations in the BRCA1 gene, our topological analysis shows that it is sensitive to the location of a mutation and the exon structure of BRCA1 in ways that cannot be found with tools based on dimensionality reduction. Moreover, the topological framework offers multiple ways to inspect results, including an error estimate that is more accurate than model uncertainty. Further studies show how these ideas produce useful results in graph-based learning and image classification.
Computer Science↗
A probabilistic graphical model foundation for enabling predictive digital twins at scale
A unifying mathematical formulation is needed to move from one-off digital twins built through custom implementations to robust digital twin implementations at scale. This work proposes a probabilistic graphical model as a formal mathematical representation of a digital twin and its associated physical asset. We create an abstraction of the asset–twin system as a set of coupled dynamical systems, evolving over time through their respective state spaces and interacting via observed data and control inputs. The formal definition of this coupled system as a probabilistic graphical model enables us to draw upon well-established theory and methods from Bayesian statistics, dynamical systems and control theory. The declarative and general nature of the proposed digital twin model make it rigorous yet flexible, enabling its application at scale in a diverse range of application areas. Here, we demonstrate how the model is instantiated to enable a structural digital twin of an unmanned aerial vehicle (UAV). The digital twin is calibrated using experimental data from a physical UAV asset. Its use in dynamic decision-making is then illustrated in a synthetic example where the UAV undergoes an in-flight damage event and the digital twin is dynamically updated using sensor data. The graphical model foundation ensures that the digital twin calibration and updating process is principled, unified and able to scale to an entire fleet of digital twins.
42 ENGINEERING↗
Theory of overparametrization in quantum neural networks
The prospect of achieving quantum advantage with quantum neural networks (QNNs) is exciting. Understanding how QNN properties (for example, the number of parameters $M$) affect the loss landscape is crucial to designing scalable QNN architectures. Here we rigorously analyze the overparametrization phenomenon in QNNs, defining overparametrization as the regime where the QNN has more than a critical number of parameters $M_c$ allowing it to explore all relevant directions in state space. In this study, our main results show that the dimension of the Lie algebra obtained from the generators of the QNN is an upper bound for $M_c$, and for the maximal rank that the quantum Fisher information and Hessian matrices can reach. Underparametrized QNNs have spurious local minima in the loss landscape that start disappearing when $M$ ≥ $M_c$. Thus, the overparametrization onset corresponds to a computational phase transition where the QNN trainability is greatly improved. We then connect the notion of overparametrization to the QNN capacity, so that when a QNN is overparametrized, its capacity achieves its maximum possible value.
97 MATHEMATICS AND COMPUTING↗
Generalizing deep learning electronic structure calculation to the plane-wave basis
Not provided.
Computer Science↗
Question-answering system extracts information on injection drug use from clinical notes
Background. Injection drug use (IDU) can increase mortality and morbidity. Therefore, identifying IDU early and initiating harm reduction interventions can benefit individuals at risk. However, extracting IDU behaviors from patients’ electronic health records (EHR) is difficult because there is no other structured data available, such as International Classification of Disease (ICD) codes, and IDU is most often documented in unstructured free-text clinical notes. Although natural language processing can efficiently extract this information from unstructured data, there are no validated tools. Methods. Here, to address this gap in clinical information, we design a question-answering (QA) framework to extract information on IDU from clinical notes for use in clinical operations. Our framework involves two main steps: (1) generating a gold-standard QA dataset and (2) developing and testing the QA model. We use 2323 clinical notes of 1145 patients curated from the US Department of Veterans Affairs (VA) Corporate Data Warehouse to construct the gold-standard dataset for developing and evaluating the QA model. We also demonstrate the QA model’s ability to extract IDU-related information from temporally out-of-distribution data. Results. Here, we show that for a strict match between gold-standard and predicted answers, the QA model achieves a 51.65% F1 score. For a relaxed match between the gold-standard and predicted answers, the QA model obtains a 78.03% F1 score, along with 85.38% Precision and 79.02% Recall scores. Moreover, the QA model demonstrates consistent performance when subjected to temporally out-of-distribution data. Conclusions. Our study introduces a QA framework designed to extract IDU information from clinical notes, aiming to enhance the accurate and efficient detection of people who inject drugs, extract relevant information, and ultimately facilitate informed patient care.
60 APPLIED LIFE SCIENCES↗
Dry electrodes with a printed cellulose–graphene ink for low-profile strain sensors in electromyography
Dihydrolevoglucosenone, commonly known as Cyrene, is a renewable and fully biodegradable cellulose-waste derived, environmentally friendly solvent, presenting a non-toxic alternative to N-methyl-2-pyrrolidone (NMP). Currently, solution-based processing of graphene and other similar van der Waals solids favor toxic solvents such as NMP, limiting their use for biosensing. However, with the use of Cyrene, bio-compatible printable devices are possible, and studies have already demonstrated its use in temperature and other biosensing methods through screen-printing. Screen-printing unfortunately often requires masks that constrain the minimum acquirable feature size to be above hundreds of centimeters and wastes material, adding to process complexity and cost. Conversely, inkjet-printing is an attractive alternative for the maskless patterning of hierarchically assembled structures, with micron length scales attainable. Graphene's high conductivity positions it ideally for long-wear sensors such as dry electrodes or respiration monitors. Here, we demonstrate the potential of Cyrene-based graphene inks through few-layer inkjet printing on flexible substrates for the first time, to produce non-toxic conductors toward a strain-mediated mechanism for biosensing, used to detect bodily motion for wearable electronics. The challenges overcome in this study include engineering ink chemistry and printing parameters such that Cyrene's relatively high viscosity compared to typical inkjet solvents, still allows for droplet ejection in a conventional material printer, yielding well-resolved clean line-edges in contrast to other solvents that exhibit diffuse line-edges possibly from stray droplets and ink-splashing. Temperature-dependent transport measurements on the inkjet-printed Cyrene-based graphene films showed the conductivity to be largely temperature-invariant but at lower temperatures below 100 K, conductivity decreased, likely as a result of increased inter-membrane separation arising from thermal contraction. Additionally, temperature-dependent Raman spectroscopy showed the red-shift in the G-band, 2D-band and D-band peaks, as temperature increased. As a result, by validating flexion motion detection of the proximal interphalangeal joint demonstrated in this study, our work is the first of its kind to successfully additively manufacture inkjet-printed Cyrene-based graphene strain sensors on flexible substrates for bio-sensing and wearables.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Chiral population analysis: a real space visualization of X-ray circular dichroism
The microscopic understanding of probing and controlling molecular chirality is of considerable interest. Numerous spectroscopic techniques are capable of monitoring molecular asymmetry and its consequences, ranging from the infrared to the X-ray regime. Resonant X-rays have long been used to investigate local atomic sites within molecules thanks to the localized nature of core electronic transitions. These techniques can be used to determine the extent to which chirality is a local versus a delocalized property. However, how to systematically partition dichroic contributions from the point of view of electronic structure simulations remains an open question. Here, we introduce the concept of chiral population analysis that connects chirality to the atomic orbital picture. In analogy with Mulliken population analysis, which assigns charges to atomic orbitals, chiral populations allow the dichroic response to be distributed among the participating atomic orbitals. This decomposition can be further visualized in real space by representing it in terms of isosurface plots, providing an intuitive way to connect the dichroic response to its origins. Thus chiral population analysis can be particularly useful to assess the extent to which a given electronic transition is sensitive to chirality as a local or global feature of the molecular geometry.
36 MATERIALS SCIENCE↗
Automated Model-Based 3D Scan Planning for Prefabricated Building Components
Not provided.
Computer Science↗
Liquid-like and rigid-body motions in molecular-dynamics simulations of a crystalline protein
To gain insight into crystalline protein dynamics, we performed molecular-dynamics (MD) simulations of a periodic 2 × 2 × 2 supercell of staphylococcal nuclease. We used the resulting MD trajectories to simulate X-ray diffraction and to study collective motions. The agreement of simulated X-ray diffraction with the data is comparable to previous MD simulation studies. We studied collective motions by analyzing statistically the covariance of alpha-carbon position displacements. The covariance decreases exponentially with the distance between atoms, which is consistent with a liquidlike motions (LLM) model, in which the protein behaves like a soft material. To gain finer insight into the collective motions, we examined the covariance behavior within a protein molecule (intraprotein) and between different protein molecules (interprotein). The interprotein atom pairs, which dominate the overall statistics, exhibit LLM behavior; however, the intraprotein pairs exhibit behavior that is consistent with a superposition of LLM and rigid-body motions (RBM). Our results indicate that LLM behavior of global dynamics is present in MD simulations of a protein crystal. They also show that RBM behavior is detectable in the simulations but that it is subsumed by the LLM behavior. Finally, the results provide clues about how correlated motions of atom pairs both within and across proteins might manifest in diffraction data. Overall, our findings increase our understanding of the connection between molecular motions and diffraction data and therefore advance efforts to extract information about functionally important motions from crystallography experiments.
59 BASIC BIOLOGICAL SCIENCES↗
Deep learning for NLTE spectral opacities
Computer simulations of high energy density science experiments are computationally challenging, consisting of multiple physics calculations including radiation transport, hydrodynamics, atomic physics, nuclear reactions, laser–plasma interactions, and more. To simulate inertial confinement fusion (ICF) experiments at high fidelity, each of these physics calculations should be as detailed as possible. However, this quickly becomes too computationally expensive even for modern supercomputers, and thus many simplifying assumptions are made to reduce the required computational time. Much of the research has focused on acceleration techniques for the various packages in multiphysics codes. In this work, we explore a novel method for accelerating physics packages via machine learning. The non-local thermodynamic equilibrium (NLTE) package is one of the most expensive calculations in the simulations of indirect drive inertial confinement fusion, taking several tens of percent of the total wall clock time. We explore the use of machine learning to accelerate this package, by essentially replacing the physics calculation with a deep neural network that has been trained to emulate the physics code. Overall, we demonstrate the feasibility of this approach on a simple problem and perform a side-by-side comparison of the physics calculation and the neural network inline in an ICF Hohlraum simulation. We show that the neural network achieves a 10× speed up in NLTE computational time while achieving good agreement with the physics code for several quantities of interest.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Experiments and simulations of isochorically heated warm dense carbon foam at the Texas Petawatt Laser
An experimental and simulation study of warm dense carbon foams at ambient density (n e ~10 21 cm –3 ) is presented. This study of isochorically heated foams is motivated by their potential application in carbon-atmosphere white-dwarf envelopes, where there are modeling uncertainties due to the equation of state. The foams are heated on an approximately picosecond time scale with a laser-accelerated proton beam. The cooling and expansion of the heated foams can be modeled with appropriately initialized radiation-hydrodynamics codes; xRAGE code is used in this work. The primary experimental diagnostic is the streaked optical pyrometer, which images a narrow band of radiation from the rear surface of the heated material. Presented are xRAGE modeling results for both solid aluminum targets and carbonized resorcinol-formaldehyde foam targets, showing that the foam appears to cool slowly on the pyrometer because of partial transparency. So that simulations of cooling foam are processed properly, it is necessary to account for finite optical depth in the photosphere calculation, and the methods for performing that calculation are presented in depth.
36 MATERIALS SCIENCE↗
Modeling laser-driven ion acceleration with deep learning
Developments in machine learning promise to ameliorate some of the challenges of modeling complex physical systems through neural-network-based surrogate models. High-intensity, short-pulse lasers can be used to accelerate ions to mega-electronvolt energies, but to model such interactions requires computationally expensive techniques such as particle-in-cell simulations. Multilayer neural networks allow one to take a relatively sparse ensemble of simulations and generate a surrogate model that can be used to rapidly search the parameter space of interest. In this study, we created an ensemble of over 1,000 simulations modeling laser-driven ion acceleration and developed a surrogate to study the resulting parameter space. A neural-network-based approach allows for rapid feature discovery not possible for traditional parameter scans given the computational cost. A notable observation made during this study was the dependence of ion energy on the pre-plasma gradient length scale. While this methodology harbors great promise for ion acceleration, it has ready application to all topics in which large-scale parameter scans are restricted by significant computational cost or relatively large, but sparse, domains.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Accuracy of the time-averaged ponderomotive approximation for laser-plasma accelerator modeling
Reliable modeling of laser-plasma accelerators, where a short and intense laser pulse propagates in an underdense plasma over long distances, is a computationally challenging task. This is due to the great disparity among the scales involved in the modeling, ranging from the micrometer scale of the laser wavelength to, for instance, the meter scale of the laser-plasma interaction length for a multi-GeV-class laser-plasma accelerator. To reduce such imbalance, the time-averaged ponderomotive approximation may be used, where the plasma particle dynamics is analytically averaged over the laser frequency, and only spatiotemporal scales associated with the laser envelope are retained in the calculations, resulting in significant computational savings. Here, we characterize the accuracy and robustness of the time-averaged ponderomotive approximation for a range of laser parameters of interest for present and future laser-plasma accelerators, and we show that the error introduced by the averaging process is small in all relevant cases.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Quadratic pseudospectrum for identifying localized states
Here we examine the utility of the quadratic pseudospectrum for understanding and detecting states that are somewhat localized in position and energy, in particular, in the context of condensed matter physics. Specifically, the quadratic pseudospectrum represents a method for approaching systems with incompatible observables {A j |1 ≤ j ≤ d} as it minimizes collectively the errors $\parallel$A j v - λ j v$\parallel$ while defining a joint approximate spectrum of incompatible observables. Moreover, we derive an important estimate relating the Clifford and quadratic pseudospectra. Finally, we prove that the quadratic pseudospectrum is local and derive the bounds on the errors that are incurred by truncating the system in the vicinity of where the pseudospectrum is being calculated.
97 MATHEMATICS AND COMPUTING↗