Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep space computing application”

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.

At least 217 records · Page 12

Validation of Loci-Stream for Autogenous Pressurization of Cryogenic Propellant Tank

Autogenous pressurization of cryogenic propellant tanks eliminates the need to have an additional pressurant tank on the space vehicle, which is highly advantageous due to reduced vehicle mass and design complexity. Autogenous pressurization therefore is one of the key technologies for deep space exploration and long-term space missions. The complex interaction of thermal gradients, turbulence and phase change near the interface make the problem a challenging one to model. Nodal analysis tools and reduced order models are unable to capture the necessary physics. 3-D CFD analyses are necessary to fully characterize autogenous pressurization. CFD analyses pose their own difficulties. The requisite CFD tool to tackle this problem need to be modular with the ability to incorporate various physics models, efficient, and computationally scalable for simulating flight size tanks. NASA MSFC's Loci-Stream CFD tool along with the VOF module is a great candidate to fit this mold. We demonstrate our modeling approach and validation of Loci-Stream for predicting autogenous pressurization of a flight scale propellant tank in order for the solver to serve as a reliable design and analysis tool for NASA's CFM application needs. Liquid hydrogen tank pressurization tests carried out at the MSFC test stand 300 facilities provide reliable validation data for this purpose. These tests were modeled using the Loci-Stream solver with a newly implemented two-phase sharp interface treatment. We show that our modeling approach and CFD solver predict the autogenous pressurization phenomena satisfactorily, and document challenging aspects of modeling this problem.

cryogenic fluid management↗

Validation of Loci-Stream for Autogenous Pressurization of Cryogenic Propellant Tank

Autogenous pressurization of cryogenic propellant tanks eliminates the need to have an additional pressurant tank on the space vehicle, which is highly advantageous due to reduced vehicle mass and design complexity. Autogenous pressurization therefore is one of the key technologies for deep space exploration and long-term space missions. The complex interaction of thermal gradients, turbulence and phase change near the interface make the problem a challenging one to model. Nodal analysis tools and reduced order models are unable to capture the necessary physics. 3-D CFD analyses are necessary to fully characterize autogenous pressurization. CFD analyses pose their own difficulties. The requisite CFD tool to tackle this problem need to be modular with the ability to incorporate various physics models, efficient, and computationally scalable for simulating flight size tanks. NASA MSFC's Loci-Stream CFD tool along with the VOF module is a great candidate to fit this mold. We demonstrate our modeling approach and validation of Loci-Stream for predicting autogenous pressurization of a flight scale propellant tank in order for the solver to serve as a reliable design and analysis tool for NASA's CFM application needs. Liquid hydrogen tank pressurization tests carried out at the MSFC test stand 300 facilities provide reliable validation data for this purpose. These tests were modeled using the Loci-Stream solver with a newly implemented two-phase sharp interface treatment. We show that our modeling approach and CFD solver predict the autogenous pressurization phenomena satisfactorily, and document challenging aspects of modeling this problem.

CFM↗

Validation of Two Phase Flow Modeling Techniques in Loci-Stream using Axial Jet Pressure Control and Autogenous Pressurization Experiment

Autogenous pressurization of cryogenic propellant tanks eliminates the need to have an additional pressurant tank on the space vehicle, which is highly advantageous due to reduced vehicle mass and design complexity. Autogenous pressurization therefore is one of the key technologies for deep space exploration and long-term space missions. The complex interaction of thermal gradients, turbulence and phase change near the interface make the problem a challenging one to model. Nodal analysis tools and reduced order models are unable to capture the necessary physics. 3-D CFD analyses are necessary to fully characterize autogenous pressurization. CFD analyses pose their own difficulties. The requisite CFD tool to tackle this problem need to be modular with the ability to incorporate various physics models, efficient, and computationally scalable for simulating flight size tanks. NASA MSFC's Loci-Stream CFD tool along with the VOF module is a great candidate to fit this mold. We demonstrate our modeling approach and validation of Loci-Stream for predicting autogenous pressurization of a flight scale propellant tank in order for the solver to serve as a reliable design and analysis tool for NASA's CFM application needs. Liquid hydrogen tank pressurization tests carried out at the MSFC test stand 300 facilities provide reliable validation data for this purpose. These tests were modeled using the Loci-Stream solver with a newly implemented two-phase sharp interface treatment. We show that our modeling approach and CFD solver predict the autogenous pressurization phenomena satisfactorily, and document challenging aspects of modeling this problem.

cryogenic fluid management↗

Innovation in Deep Space Habitat Interior Design: Lessons Learned From Small Space Design in Terrestrial Architecture

Increased public awareness of carbon footprints, crowding in urban areas, and rising housing costs have spawned a 'small house movement' in the housing industry. Members of this movement desire small, yet highly functional residences which are both affordable and sensitive to consumer comfort standards. In order to create comfortable, minimum-volume interiors, recent advances have been made in furniture design and approaches to interior layout that improve both space utilization and encourage multi-functional design for small homes, apartments, naval, and recreational vehicles. Design efforts in this evolving niche of terrestrial architecture can provide useful insights leading to innovation and efficiency in the design of space habitats for future human space exploration missions. This paper highlights many of the cross-cutting architectural solutions used in small space design which are applicable to the spacecraft interior design problem. Specific solutions discussed include reconfigurable, multi-purpose spaces; collapsible or transformable furniture; multi-purpose accommodations; efficient, space saving appliances; stowable and mobile workstations; and the miniaturization of electronics and computing hardware. For each of these design features, descriptions of how they save interior volume or mitigate other small space issues such as confinement stress or crowding are discussed. Finally, recommendations are provided to provide guidance for future designs and identify potential collaborations with the small spaces design community.

Simon, Matthew A.↗

Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations

Here we propose a generalized space-time domain decomposition approach for the physics-informed neural networks (PINNs) to solve nonlinear partial differential equations (PDEs) on arbitrary complex-geometry domains. The proposed framework, named eXtended PINNs ( X P I N N s ), further pushes the boundaries of both PINNs as well as conservative PINNs (cPINNs), which is a recently proposed domain decomposition approach in the PINN framework tailored to conservation laws. Compared to PINN, the XPINN method has large representation and parallelization capacity due to the inherent property of deployment of multiple neural networks in the smaller subdomains. Unlike cPINN, XPINN can be extended to any type of PDEs. Moreover, the domain can be decomposed in any arbitrary way (in space and time), which is not possible in cPINN. Thus, XPINN offers both space and time parallelization, thereby reducing the training cost more effectively. In each subdomain, a separate neural network is employed with optimally selected hyperparameters, e.g., depth/width of the network, number and location of residual points, activation function, optimization method, etc. A deep network can be employed in a subdomain with complex solution, whereas a shallow neural network can be used in a subdomain with relatively simple and smooth solutions. We demonstrate the versatility of XPINN by solving both forward and inverse PDE problems, ranging from one-dimensional to three-dimensional problems, from time-dependent to time-independent problems, and from continuous to discontinuous problems, which clearly shows that the XPINN method is promising in many practical problems. The proposed XPINN method is the generalization of PINN and cPINN methods, both in terms of applicability as well as domain decomposition approach, which efficiently lends itself to parallelized computation. The XPINN code is available on h t t p s : / / g i t h u b . c o m / A m e y a J a g t a p / X P I N N s .

97 MATHEMATICS AND COMPUTING↗

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↗

Deep learning of parameterized equations with applications to uncertainty quantification

We propose a learning algorithm for discovering unknown parameterized dynamical systems by using observational data of the state variables. Our method is built upon and extends the recent work of discovering unknown dynamical systems, in particular those using deep neural network (DNN). We propose a DNN structure, largely based upon the residual network (ResNet), to not only learn the unknown form of the governing equation but also take into account the random effect embedded in the system, which is generated by the random parameters. Once the DNN model is successfully constructed, it is able to produce system prediction over longer term and for arbitrary parameter values. For uncertainty quantification, it allows us to conduct uncertainty analysis by evaluating solution statistics over the parameter space.

97 MATHEMATICS AND COMPUTING↗

INTEGRATE - Inverse Network Transformations for Efficient Generation of Robust Airfoil and Turbine Enhancements

The INTEGRATE (Inverse Network Transformations for Efficient Generation of Robust Airfoil and Turbine Enhancements) project is developing a new inverse-design capability for the aerodynamic design of wind turbine rotors using invertible neural networks. This AI-based design technology can capture complex non-linear aerodynamic effects while being 100 times faster than design approaches based on computational fluid dynamics. This project enables innovation in wind turbine design by accelerating time to market through higher-accuracy early design iterations to reduce the levelized cost of energy. INVERTIBLE NEURAL NETWORKS Researchers are leveraging a specialized invertible neural network (INN) architecture along with the novel dimension-reduction methods and airfoil/blade shape representations developed by collaborators at the National Institute of Standards and Technology (NIST) learns complex relationships between airfoil or blade shapes and their associated aerodynamic and structural properties. This INN architecture will accelerate designs by providing a cost-effective alternative to current industrial aerodynamic design processes, including: - Blade element momentum (BEM) theory models: limited effectiveness for design of offshore rotors with large, flexible blades where nonlinear aerodynamic effects dominate - Direct design using computational fluid dynamics (CFD): cost-prohibitive - Inverse-design models based on deep neural networks (DNNs): attractive alternative to CFD for 2D design problems, but quickly overwhelmed by the increased number of design variables in 3D problems AUTOMATED COMPUTATIONAL FLUID DYNAMICS FOR TRAINING DATA GENERATION - MERCURY FRAMEWORK The INN is trained on data obtained using the University of Marylands (UMD) Mercury Framework, which has with robust automated mesh generation capabilities and advanced turbulence and transition models validated for wind energy applications. Mercury is a multi-mesh paradigm, heterogeneous CPU-GPU framework. The framework incorporates three flow solvers at UMD, 1) OverTURNS, a structured solver on CPUs, 2) HAMSTR, a line based unstructured solver on CPUs, and 3) GARFIELD, a structured solver on GPUs. The framework is based on Python, that is often used to wrap C or Fortran codes for interoperability with other solvers. Communication between multiple solvers is accomplished with a Topology Independent Overset Grid Assembler (TIOGA). NOVEL AIRFOIL SHAPE REPRESENTATIONS USING GRASSMAN SPACES We developed a novel representation of shapes which decouples affine-style deformations from a rich set of data-driven deformations over a submanifold of the Grassmannian. The Grassmannian representation as an analytic generative model, informed by a database of physically relevant airfoils, offers (i) a rich set of novel 2D airfoil deformations not previously captured in the data , (ii) improved low-dimensional parameter domain for inferential statistics informing design/manufacturing, and (iii) consistent 3D blade representation and perturbation over a sequence of nominal shapes. TECHNOLOGY TRANSFER DEMONSTRATION - COUPLING WITH NREL WISDEM Researchers have integrated the inverse-design tool for 2D airfoils (INN-Airfoil) into WISDEM (Wind Plant Integrated Systems Design and Engineering Model), a multidisciplinary design and optimization framework for assessing the cost of energy, as part of tech-transfer demonstration. The integration of INN-Airfoil into WISDEM allows for the design of airfoils along with the blades that meet the dynamic design constraints on cost of energy, annual energy production, and the capital costs. Through preliminary studies, researchers have shown that the coupled INN-Airfoil + WISDEM approach reduces the cost of energy by around 1% compared to the conventional design approach. This page will serve as a place to easily access all the publications from this work and the repositories for the software developed and released through this pr...

aerodynamics↗

Latent Space Simulation for Carbon Capture Design Optimization

The CO2 capture efficiency in solvent-based carbon capture systems (CCSs) critically depends on the gas-solvent interfacial area (IA), making maximization of IA a foundational challenge in CCS design. While the IA associated with a particular CCS design can be estimated via a computational fluid dynamics (CFD) simulation, using CFD to derive the IAs associated with numerous CCS designs is prohibitively costly. However, previous works such as Deep Fluids (Kim et al., 2019) show that large simulation speed-ups and low error are achievable by replacing CFD simulators with neural-network (NN) surrogates that mimic the CFD simulation process. This raises the possibility for a fast, accurate replacement for a CFD simulator, and thus computationally feasible IA-based CCS-design optimization. As such, here, we explore whether an existing NN-surrogate approach (and variants we develop) can successfully be applied to our complex carbon-capture CFD simulations, with the ultimate goal of obtaining a fast and accurate simulator for our CCS-design application. Our experiments build on the Deep Fluids approach and find that resulting surrogates can produce large speed ups (4000x) while maintaining IA relative errors as low as 4% on unseen CCS configurations (interpolating between configurations seen during training). Thus, despite less faithfulness to the underlying physics of the problem, NN surrogates may be a promising tool for our CCS design optimization problem. Notably, though, the Deep Fluids approach has limitations for our application (such as model non-transferability to CCS-packing changes), which we discuss. We conclude with potential directions for future work that may, like innovations we introduced here (e.g., transformer-based dynamics prediction), improve performance on our complicated dataset of CCS CFD simulations.

AI Surrogate Simulation, Carbon Capture, design op↗

Gravitational wave signatures of the absence of an event horizon: Nonradial oscillations of a thin-shell gravastar

Gravitational waves from compact objects provide information about their structure, probing deep into strong-gravity regions. Here we illustrate how the presence or absence of an event horizon can produce qualitative differences in the gravitational waves emitted by ultracompact objects. In order to set up a straw-man ultracompact object with no event horizon, but which is otherwise almost identical to a black hole, we consider a nonrotating thin-shell model inspired by Mazur and Mottola's gravastar, which has a Schwarzschild exterior, a de Sitter interior and an infinitely thin shell with finite tension separating the two regions. As viewed from the external space-time, the shell can be located arbitrarily close to the Schwarzschild radius, so a gravastar might seem indistinguishable from a black hole when tests are only performed on its external metric. We study the linearized dynamics of the system, and, in particular, the junction conditions connecting internal and external gravitational perturbations. As a first application of the formalism we compute polar and axial oscillation modes of a thin-shell gravastar. We show that the quasinormal mode spectrum is completely different from that of a black hole, even in the limit when the surface redshift becomes infinite. Polar quasinormal modes depend on the equation of state of matter on the shell and can be used to distinguish between different gravastar models. Our calculations suggest that low-compactness gravastars could be unstable when the sound speed on the shell v{sub s}/c > or approx. 0.92.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Simulating Atmospheric Processes in Earth System Models and Quantifying Uncertainties With Deep Learning Multi‐Member and Stochastic Parameterizations

Abstract Deep learning is a powerful tool to represent subgrid processes in climate models, but many application cases have so far used idealized settings and deterministic approaches. Here, we develop stochastic parameterizations with calibrated uncertainty quantification to learn subgrid convective and turbulent processes and surface radiative fluxes of a superparameterization embedded in an Earth System Model (ESM). We explore three methods to construct stochastic parameterizations: (a) a single Deep Neural Network (DNN) with Monte Carlo Dropout; (b) a multi‐member parameterization; and (c) a Variational Encoder Decoder with latent space perturbation. We show that the multi‐member parameterization improves the representation of convective processes, especially in the planetary boundary layer, compared to individual DNNs. The respective uncertainty quantification illustrates that methods (b) and (c) are advantageous compared to a dropout‐based DNN parameterization regarding the spread of convective processes. Hybrid simulations with our best‐performing multi‐member parameterizations remained challenging and crash within the first days. Therefore, we develop a pragmatic partial coupling strategy relying on the superparameterization for condensate emulation. Partial coupling reduces the computational efficiency of hybrid Earth‐like simulations but enables model stability over 5 months with our multi‐member parameterizations. However, our hybrid simulations exhibit biases in thermodynamic fields and differences in precipitation patterns. Despite this, the multi‐member parameterizations enable improvements in reproducing tropical extreme precipitation compared to a traditional convection parameterization. Despite these challenges, our results indicate the potential of a new generation of multi‐member machine learning parameterizations leveraging uncertainty quantification to improve the representation of stochasticity of subgrid effects.

Behrens, Gunnar [Deutsches Zentrum für Luft‐ und R↗

Union: A Unified HW-SW Co-Design Ecosystem in MLIR for Evaluating Tensor Operationson Spatial Accelerators

To meet the extreme compute demands for deep learning across commercial and scientific applications, dataflow accelerators are becoming increasingly popular. While these“domain-specific” accelerators are not fully programmable like CPUs and GPUs, they retain varying levels of flexibility with respect to data orchestration, i.e., dataflow and tiling optimizations to enhance efficiency. There are several challenges when designing new algorithms and mapping approaches to execute the algorithms for a target problem on new hardware. Previous works have addressed these challenges individually. To address this challenge as a whole, in this work, we present an HW-SW co-design ecosystem for spatial accelerators called Union within the popular MLIR compiler infrastructure. Our framework allows exploring different algorithms and their mappings on several accelerator cost models. Union also includes a plug-and-play library of accelerator cost models and mappers which can easily be extended. The algorithms and accelerator cost models are connected via a novel mapping abstraction that captures the map space of spatial accelerators which can be systematically pruned based on constraints from the hardware, workload, and mapper. We demonstrate the value of Union for the community with several case studies which examine offloading different tensor operations (CONV/GEMM/Tensor Contraction) on diverse accelerator architectures using different mapping schemes.

Jeong, Geonhwa↗

Electromagnetic deep-probing (100-1000 kms) of the Earth's interior from artificial satellites: Constraints on the regional emplacement of crustal resources

A spherical harmonic analysis program is being tested which takes magnetic data in universal time from a set of arbitrarily space observatories and calculates a value for the instantaneous magnetic field at any point on the globe. The calculation is done as a least mean-squares value fit to a set of spherical harmonics up to any desired order. The program accepts as a set of input the orbit position of a satellite coordinates it with ground-based magnetic data for a given time. The output is a predicted time series for the magnetic field on the Earth's surface at the (r, theta) position directly under the hypothetically orbiting satellite for the duration of the time period of the input data set. By tracking the surface magnetic field beneath the satellite, narrow-band averages crosspowers between the spatially coordinated satellite and the ground-based data sets are computed. These crosspowers are used to calculate field transfer coefficients with minimum noise distortion. The application of this technique to calculating the vector response function W is discussed.

Hermance, J. F.↗

Bayesian sparse learning with preconditioned stochastic gradient MCMC and its applications

Deep neural networks have been successfully employed in an extensive variety of research areas, including solving partial differential equations. Despite its significant success, there are some challenges in effectively training DNN, such as avoiding overfitting in over-parameterized DNNs and accelerating the optimization in DNNs with pathological curvature. Here, we propose a Bayesian type sparse deep learning algorithm. The algorithm utilizes a set of spike-and-slab priors for the parameters in the deep neural network. The hierarchical Bayesian mixture will be trained using an adaptive empirical method. That is, one will alternatively sample from the posterior using preconditioned stochastic gradient Langevin Dynamics (PSGLD), and optimize the latent variables via stochastic approximation. The sparsity of the network is achieved while optimizing the hyperparameters with adaptive searching and penalizing. A popular SG-MCMC approach is Stochastic gradient Langevin dynamics (SGLD). However, considering the complex geometry in the model parameter space in nonconvex learning, updating parameters using a universal step size in each component as in SGLD may cause slow mixing. To address this issue, we apply a computationally manageable preconditioner in the updating rule, which provides a step-size parameter to adapt to local geometric properties. Moreover, by smoothly optimizing the hyperparameter in the preconditioning matrix, our proposed algorithm ensures a decreasing bias, which is introduced by ignoring the correction term in the preconditioned SGLD. According to the existing theoretical framework, we show that the proposed algorithm can asymptotically converge to the correct distribution with a controllable bias under mild conditions. Numerical tests are performed on both synthetic regression problems and learning solutions of elliptic PDE, which demonstrate the accuracy and efficiency of the present work.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Neural network representations of multiphase Equations of State

Abstract Equations of State model relations between thermodynamic variables and are ubiquitous in scientific modelling, appearing in modern day applications ranging from Astrophysics to Climate Science. The three desired properties of a general Equation of State model are adherence to the Laws of Thermodynamics, incorporation of phase transitions, and multiscale accuracy. Analytic models that adhere to all three are hard to develop and cumbersome to work with, often resulting in sacrificing one of these elements for the sake of efficiency. In this work, two deep-learning methods are proposed that provably satisfy the first and second conditions on a large-enough region of thermodynamic variable space. The first is based on learning the generating function (thermodynamic potential) while the second is based on structure-preserving, symplectic neural networks, respectively allowing modifications near or on phase transition regions. They can be used either “from scratch” to learn a full Equation of State, or in conjunction with a pre-existing consistent model, functioning as a modification that better adheres to experimental data. We formulate the theory and provide several computational examples to justify both approaches, highlighting their advantages and shortcomings.

Science & Technology - Other Topics↗

Attribute-Guided Adversarial Training for Robustness to Natural Perturbations

We report while existing work in robust deep learning has focused on small pixel-level norm-based perturbations, this may not account for perturbations encountered in several real world settings. In many such cases although test data might not be available, broad specifications about the types of perturbations (such as an unknown degree of rotation) may be known. We consider a setup where robustness is expected over an unseen test domain that is not i.i.d. but deviates from the training domain. While this deviation may not be exactly known, its broad characterization is specified a priori, in terms of attributes. We propose an adversarial training approach which learns to generate new samples so as to maximize exposure of the classifier to the attributes-space, without having access to the data from the test domain. Our adversarial training solves a min-max optimization problem, with the inner maximization generating adversarial perturbations, and the outer minimization finding model parameters by optimizing the loss on adversarial perturbations generated from the inner maximization. We demonstrate the applicability of our approach on three types of naturally occurring perturbations --- object-related shifts, geometric transformations, and common image corruptions. Our approach enables deep neural networks to be robust against a wide range of naturally occurring perturbations. We demonstrate the usefulness of the proposed approach by showing the robustness gains of deep neural networks trained using our adversarial training on MNIST, CIFAR-10, and a new variant of the CLEVR dataset.

97 MATHEMATICS AND COMPUTING↗

Computation on collisionless steady-state plasma flow past a charged disk

A computer method is presented using the 'inside-out' approach, for predicting the structure of the disturbed zone near a moving body in space. The approach uses fewer simplifying assumptions than other available methods, and is applicable to large ranges of the values of body and plasma parameters. Two major advances concerning 3-dimensional bodies are that thermal motions of ions as well as of electrons are treated realistically by following their trajectories in the electric field, and the technique for achieving self-consistency is promising for very large bodies. Three sample solutions were obtained for a disk-shaped body, charged negatively to a potential 4kT/e. With ion Mach number 4, and equal ion and electron temperatures, the wakes of a relatively small body (radius 5 Debye lengths) and a relatively large body (radius 100 Debye lengths) both begin to fill up between 2 and 3 body radii downstream. For the large body there is in addition a potential well (about 6kT/e deep) behind the body. Increasing the ion Mach number to 8 for the large body causes the potential well to become wider and longer but not deeper. For the large body, the quasineutrality assumption is validated outside of a cone-shaped region in the very near wake. For the large as well as the small body, the disturbed zone behind the body extends transversely no more than 2 or 3 body radii, a result of significance for the design of spacecraft boom instrumentation.

Parker, L. W.↗

LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields

Model ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of per-sample errors obtained during training and employing a distance-based similarity search in the model latent space. Our method, which we call LTAU (Loss Trajectory Analysis for Uncertainty), efficiently estimates the full probability distribution function of errors for any test point using the logged training errors, achieving speeds that are 2–3 orders of magnitudes faster than typical ensemble methods and allowing it to be used for tasks where training or evaluating multiple models would be infeasible. We apply LTAU towards estimating parametric uncertainty in atomistic force fields (LTAU-FF), demonstrating that it produces well-calibrated confidence intervals and predicts errors that correlate strongly with the true errors for data near the training domain. Furthermore, we show that the errors predicted by LTAU-FF can be used in practical applications for detecting out-of-domain data, tuning model performance, and predicting failure during simulations. We believe that LTAU will be a valuable tool for uncertainty quantification in atomistic force fields and is a promising method that should be further explored in other domains of machine learning.

97 MATHEMATICS AND COMPUTING↗