SEARCH · Search NASA

Results for “Gradient estimation”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5

All sources NASA NTRS DOE OSTI

DOE OSTI2024

AEPF: Attention-Enabled Point Fusion for 3D Object Detection

Current state-of-the-art (SOTA) LiDAR-only detectors perform well for 3D object detection tasks, but point cloud data are typically sparse and lacks semantic information. Detailed semantic information obtained from camera images can be added with existing LiDAR-based detectors to create a robust 3D detection pipeline. With two different data types, a major challenge in developing multi-modal sensor fusion networks is to achieve effective data fusion while managing computational resources. With separate 2D and 3D feature extraction backbones, feature fusion can become more challenging as these modes generate different gradients, leading to gradient conflicts and suboptimal convergence during network optimization. To this end, we propose a 3D object detection method, Attention-Enabled Point Fusion (AEPF). AEPF uses images and voxelized point cloud data as inputs and estimates the 3D bounding boxes of object locations as outputs. An attention mechanism is introduced to an existing feature fusion strategy to improve 3D detection accuracy and two variants are proposed. These two variants, AEPF-Small and AEPF-Large, address different needs. AEPF-Small, with a lightweight attention module and fewer parameters, offers fast inference. AEPF-Large, with a more complex attention module and increased parameters, provides higher accuracy than baseline models. Experimental results on the KITTI validation set show that AEPF-Small maintains SOTA 3D detection accuracy while inferencing at higher speeds. AEPF-Large achieves mean average precision scores of 91.13, 79.06, and 76.15 for the car class’s easy, medium, and hard targets, respectively, in the KITTI validation set. Results from ablation experiments are also presented to support the choice of model architecture.

Chemistry↗

DOE OSTI2026

Temperature profiling at the American WAKE ExperimeNt (AWAKEN): methodology and uncertainty quantification

We quantify the accuracy of the temperature profiling from ground-based spectral infrared radiance observations at the American WAKE ExperimeNt (AWAKEN). Results from pre-campaign tests and comparisons with in-situ ground-based and airborne sensors at AWAKEN indicate that temperature profiles agree satisfactorily with traditional instruments for wind energy applications. The bias is within a fraction of a degree and appears to be related to atmospheric stability. Root-mean-square differences from the reference instruments are always smaller than a degree and are often well described by the online uncertainty estimation product. Height-to-height and site-to-site temperature differences are in excellent agreement with in-situ observations, which justifies the use of temperature profilers to characterize static stability and spatial gradients of temperature.

17 WIND ENERGY↗

DOE OSTI2025

Reproductive and leaf litterfall fluxes in forest ecosystem sites globally (1950-2022)

Forest allocation of net primary productivity (NPP) to reproduction is poorly quantified globally, despite its critical role in forest regeneration and a well-supported trade-off with allocation to growth. Although field measurements of total NPP are rare, our work finds that a proxy for reproductive carbon allocation constructed from leaf (L) and reproductive (R) litterfall fluxes, R/(R+L), is strongly correlated with R/NPP, facilitating analysis across a wide range of sites where biometric estimates of NPP are not available (R² = 0.85; Hanbury-Brown et al., 2022, Ward et al., in prep). To investigate relationships between ecosystem-scale reproductive allocation (RA) and climate, soil fertility, and stand age gradients, we conducted a literature search and synthesized 824 observations of annual average leaf and reproductive litterfall fluxes across forest sites globally. The zip file includes 1) a folder Data/ containing the litterfall data ("GlobalForestRA_data.csv") and metadata ("GlobalForestRA_metadata.doc") files. The data file includes geographic coordinates, long-term mean annual temperature and precipitation (1970-2000, extracted from WorldClim2.1), leaf and reproductive litterfall fluxes, sampling interval and protocols, forest characteristics (dominant leaf morphology, information pertaining to forest age and successional stage, and disturbance history) and soil properties (% sand, %silt, %clay, total phosphorus (P), nitrogen (N), cation exchange capacity (CEC) and pH) extracted from SoilGrids250 and from on-site measurements, where available. The metadata file contains information about each variable reported in the data file, including data sources, processing methods, and all references. The Data folder contains two additional files used to create Figure 1; these are described in greater detail in the README.2) R scripts GloalForestRA_analysis.r and GlobalForestRA_SI.r and a folder /Functions used to produce results, figures, and tables in the manuscript Ward et al. (in press)3) a README file describing how the data and R scripts can be used to reproduce statistical results, figures, and tables found in the manuscript. Ward et al. (in press)This repository can also be found at: https://github.com/r-ward/Global_Analysis_ForestRA.Ward, R.E., Zhang-Zheng, H. Aernethy, K., Adu-Bredu, S., Arroyo, L., Bailey, A. et al. (in press). Forest age rivals climate to explain reproductive allocation patterns in forest ecosystems globally. Ecology Letters. Hanbury-Brown, A.R., Ward, R.E. & Kueppers, L.M. (2022). Forest regeneration within Earth system models: current process representations and ways forward. New Phytol., 235, 20–40.Ward et al. (2025), Forest age rivals climate to explain reproductive allocation patterns in forest ecosystems globally, in prep.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Global particle buildup simulations with gas puff scan: application to WEST discharge

This paper deals with the distribution of sources, transport, and exhaust of particles in a tokamak. Knowledge and understanding of all the physical phenomena involved in the global particle buildup are necessary to study and predict density regimes and subsequently to develop optimized scenarios for tokamak operation in order to control heat and particle exhaust. Neutral particles and their interactions with plasma are central in this perspective. This paper discusses the impact of varying the intensity of particle fueling in 2D transport simulations of a WEST discharge. Simulations are performed with an updated version of SOLEDGE-HDG that allows a more realistic transport of neutrals using a self-consistent diffusive model based on charge exchange and ionization processes. New code capabilities allow the entire WEST poloidal cross section to be simulated in a realistic configuration for both geometry and the range of control parameters. A gas puff scan illustrates the main features of the sheath-limited, high-recycling, and detached regimes, such as the buildup of the temperature gradient and the pressure drop in the scrape-off layer (SOL), the target temperature falling to 1 eV, and the ionization source moving away from the targets, as well as the particle flux rollover. A crude estimate of wall erosion is also provided, showing the respective role of each plasma wall component in each of these regimes.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2024

Characterizing and improving the performance of molten-salt-steam heat exchangers in concentrating solar power plants

Shell-and-tube heat exchangers (HXs) for steam generation from molten salts in concentrating solar power (CSP) plants experience thermal fatigue due to significant temperature gradients and inherent transient operation. Molten salt-steam HX design lifespans exceed actual lifespans, and, as a consequence, designers overpredict plant profitability and operators neglect appropriate prescriptions to optimize these lifetimes. Here, this study refines HX lifespan estimates with data benchmarked against thermal-fluid mechanical modeling of stress and accumulated fatigue. Reduced-order thermal models of the molten salt-steam, shell-and-tube evaporator and superheater predict transient temperature profiles along the two HXs salt-steam flow paths. The modeled evaporator and superheater temperature profiles enable assessment of cyclic stresses within the HX tubesheets, where molten-salt HX failures are most common. Evaporator and superheater performance data from a current 110 MW elec commercial CSP plant provide a basis for validating the reduced-order HX models. HX life predictions derived from stochastic failure distributions serve as inputs for simulating and optimizing existing plant operations. The impact of the updated lifespans on overall plant revenue depends on operating scenarios. This study suggests that typical ramping rates for a CSP plant with a high-temperature Rankine cycle result in an evaporator and superheater life of approximately 10 and 25 years, respectively, compared to the design target of 30 years. Reduced HX lifespans decrease operational plant revenue on average by 4.6-5.1%. Furthermore, there may be as many as four HX replacements over the 30-year lifetime of the plant; and, purchase agreement loss due to failure to meet contractual production requirements can have ramifications that include the risk of bankruptcy.

14 SOLAR ENERGY↗

DOE OSTI2025

Bottom-Up Simulation, Reconstruction, and Quantification of Macromolecule Sequences from Experimental Polymerizations

Motivated by the canonical sequence–structure–function paradigm, tools to characterize chemical patterning in natural biomacromolecules, from proteins to nucleic acids, have grown exponentially in recent years. However, analogous strategies for synthetic macromolecules remain in nascent stages, complicated by sequence polydispersity and analytical limitations. To address this, we have developed a comprehensive and open-source Python package, PRISM (polymer rate insights and sequence modeling), an end-to-end workflow that provides a path from experimental kinetics measurements to quantitative and qualitative metrics for describing chemical patterning in stochastic polymers. First, a numerical integration strategy was constructed to simulate and fit experimental data from reversible addition–fragmentation chain transfer (RAFT) polymerization kinetics, enabling the facile estimation of relevant reactivity ratios. These ratios were then used in a mechanism-specific stochastic kinetic simulation strategy to simulate sequence ensembles corresponding to model systems spanning experimental copolymers, classes of statistical polymers (e.g., alternating, block, and gradient), and multiblock copolymers. Lastly, inspired by sequence homology metrics from bioinformatics, we introduce visualization strategies and quantitative metrics to facilitate comparisons of different sequence ensembles. As the sequence–structure–function paradigm becomes increasingly central in de novo design of synthetic macromolecules, this toolkit provides a first step toward accurate and representative sequence description and featurization.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2025

Scalable Bayesian Physics-Informed Kolmogorov-Arnold Networks

Uncertainty quantification (UQ) plays a pivotal role in scientific machine learning, especially when surrogate models are used to approximate complex systems. Although multilayer perceptions (MLPs) are commonly employed as surrogates, they often suffer from overfitting due to their large number of parameters. Kolmogorov-Arnold networks (KANs) offer an alternative solution with fewer parameters. However, gradient-based inference methods, such as Hamiltonian Monte Carlo (HMC), may result in computational inefficiency when applied to KANs, especially for large-scale datasets, due to the high cost of back-propagation. To address these challenges, we propose a novel approach, combining the dropout Tikhonov ensemble Kalman inversion (DTEKI) with Chebyshev KANs. This gradient-free method effectively mitigates overfitting and enhances numerical stability. In addition, we incorporate the active subspace method to reduce the parameter-space dimensionality, allowing us to improve the accuracy of predictions and obtain more reliable uncertainty estimates. Extensive experiments demonstrate the efficacy of our approach in various test cases, including scenarios with large datasets and high noise levels. Our results show that the new method achieves comparable or better accuracy, much higher efficiency as well as stability compared to HMC, in addition to scalability. Moreover, by leveraging the low-dimensional parameter subspace, our method preserves prediction accuracy while substantially reducing further the computational cost.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

CO 2 storage site characterization using ensemble-based approaches with deep generative models

Estimating spatially distributed properties such as permeability from available sparse measurements is a great challenge in efficient subsurface CO 2 storage operations. In this paper, a deep generative model that can accurately capture complex subsurface structure is tested with an ensemble-based inversion method for accurate and accelerated characterization of CO 2 storage sites. We chose Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) for its realistic reservoir property representation and Ensemble Smoother with Multiple Data Assimilation (ES-MDA) for its robust data fitting and uncertainty quantification capability. WGAN-GP are trained to generate high-dimensional permeability fields from a low-dimensional latent space and ES-MDA then updates the latent variables by assimilating available measurements. Several subsurface site characterization examples including Gaussian, channelized, and fractured reservoirs are used to evaluate the accuracy and computational efficiency of the proposed method and the main features of the unknown permeability fields are characterized accurately with reliable uncertainty quantification. Furthermore, the estimation performance is compared with a widely-used variational, i.e., optimization-based, inversion approach, and the proposed approach outperforms the variational inversion method in several benchmark cases. We explain such superior performance by visualizing the objective function in the latent space: because of nonlinear and aggressive dimension reduction via generative modeling, the objective function surface becomes extremely complex while the ensemble approximation can smooth out the multi-modal surface during the minimization. This suggests that the ensemble-based approach works well over the variational approach when combined with deep generative models at the cost of forward model runs unless convergence-ensuring modifications are implemented in the variational inversion.

42 ENGINEERING↗

DOE OSTI2025

Water Stable Isotopes in Precipitation, Rivers, and Groundwater Across an Elevation Gradient in the Sierra Nevada Mountains (USA) Reflect Source Elevation

Understanding watershed processes is critical to predict the impacts of climate change and forest management on water resources. However, collecting hydrological data in mountainous terrain is challenging. Precipitation, river water, and groundwater H and O stable isotope data can provide insights into processes occurring at the mountain range scale. Water δ 2 H and δ 18 O values in precipitation vary with terrain elevation; thus, the resulting isotopic lapse rates of precipitation, groundwater, and river water have the potential to elucidate watershed processes and source elevations of major rivers. We analysed H and O stable isotope data of precipitation, groundwater, and river water over the course of one Water Year (Oct 2016—Oct 2017) in the Sierra Nevada mountains of California, USA. We calculated elevation-dependent isotopic lapse rates of these waters to estimate the source elevation of major rivers draining the west flank of the Sierra Nevada mountains. We also investigated the Cosumnes River's watershed in more detail to determine how river flow may be more fully partitioned. Here, we found that H and O stable isotopes in precipitation are temporally variable, but isotopic lapse rates are generally consistent with prior studies. However, groundwater samples across an elevation gradient provide a more consistent and accessible isotopic lapse rate to predict river water source elevations.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2025

A data-driven method to estimate the antiproton background in the Mu2e experiment

The Mu2e experiment at Fermilab will search for the Charged Lepton Flavour Violating (CLFV) process of coherent, neutrinoless µ− → e − conversion in the field of an aluminum nucleus. The expected signal is a monochromatic electron with the energy of 104.97 MeV, slightly below the muon rest mass. Observation of a CLFV process would provide unambiguous evidence for Beyond the Standard Model (BSM) physics. Mu2e is sensitive to a wide range of BSM models and has the capability to distinguish between them, guiding us towards the most accurate models. The key features of the Mu2e experiment are: (1) a high intensity pulsed negative muon beam with about 1010 stopped µ −/s, and (2) a sophisticated superconducting solenoid system with a gradient magnetic field to form and guide the intense muon beam to the target. The Mu2e physics data taking is expected to begin in 2027. For Run I, the expected 5σ discovery sensitivity is Rµe = 1.2 × 10−15, with a total expected background of 0.11 ± 0.03 events. In the absence of a signal, the expected upper limit is Rµe < 6.2 × 10−16 at 90% CL. The success of this experiment hinges on the accurate estimation of the background from various SM processes that could provide signal-like electrons. One of the background processes is antiprotons annihilating in the stopping target to produce signal like electrons through π0 → γγ decays followed by γ conversions, and π− → µ−ν¯ decays followed by µ− decay. It is a relatively small background with large uncertainty (100%) due to the lack of antiproton production cross section information for the Mu2e proton beam energy of 8 GeV. We have developed a novel methodology to estimate the antiproton background in-situ. This forms the main theme of the thesis. We observed that at Mu2e energies, antiproton annihilation in the stopping target is the only source of events with multiple, simultaneous particle trajectories. From Geant4 simulations, only about 0.2% of the simulated antiproton annihilation events have a signal-like electron. Meanwhile, ∼ 5% of events have multiple reconstructible particle tracks per event. Therefore, we have devised a methodology to reconstruct the multi-track events and estimate the antiproton background by exploiting the large ratio of the production rates of the two final states.

Chithirasreemadam, Namitha [Pisa U.] (ORCID:000000↗

DOE OSTI2026

Data from: Understanding the biogeochemical and spatial drivers of methane and carbon dioxide fluxes in a large temperate reservoir

This dataset contains spatially resolved measurements of CO₂ and CH₄ fluxes and associated environmental variables collected across 200 sites in Douglas Reservoir (Tennessee, USA) between July 29-August 2, 2024. Measurements include diffusive fluxes of CO₂ and CH₄, CH₄ ebullition, and biogeochemical and spatial variables such as dissolved oxygen, temperature, conductivity, chlorophyll-a, pH, water depth, and distance from the dam. Sampling was conducted using a spatially balanced design to capture longitudinal and depth-related gradients throughout the reservoir. The dataset is structured to support analyses of spatial variability, flux pathway comparisons, and modeling approaches (e.g., spatial statistics and machine learning) aimed at understanding controls on reservoir CO₂ and CH₄ fluxes and improving upscaling to whole-reservoir and regional estimates.

Neeper, Jamie [ORNL] (ORCID:0009000842342101)↗

DOE OSTI2025

CFD modeling of natural circulation in LiCl-KCl molten salt closed loop

Characterizing flow within a molten salt closed-loop system is crucial for assessing system requirements, evaluating performance, and identifying potential flaws. Direct flow measurement using instrumentation is challenging due to extreme environmental conditions and the limitations associated with measuring molten salt flow under natural convection. Here, this study aims to provide comprehensive insights into the thermal-hydraulic behavior of a closed loop, with a particular focus on temperature distribution and velocity prediction. The Computational Fluid Dynamics (CFD) model demonstrated the capability to effectively simulate and predict both temperature distributions and flow velocities within the molten salt loop. The CFD model's predictive capability was validated by its ability to replicate temperature measurements under varying boundary conditions. The analysis revealed that the CFD model tends to underpredict temperatures in the cold leg and overpredict them in the hot leg, highlighting the need for continuous model refinement and acknowledging the limitations of using a steady-state approach. Furthermore, the potential of using external temperature measurements to estimate internal molten salt temperatures and predict flow velocity was explored, revealing that this approach could introduce up to a 5.5% error in flow velocity calculations. Line probes mapping temperature distributions across the tube's cross-section and molten salt provided valuable insights into temperature gradients, emphasizing the need for a thermal conductivity equation for molten salt with lower uncertainty to achieve more accurate temperature predictions of the system.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2026

Bayesian Exploration and Surrogate Emulation of Nonlinear Beam-Response Geometry in the LBNF Beamline

Next-generation long-baseline neutrino experiments aim to achieve multi-MW proton beam power while reducing accelerator-induced systematic uncertainties. At Fermilab, the LBNF beamline is designed for 1.2 MW operation with PIP-II and is upgradeable to 2.4 MW. DUNE will probe the three-flavor neutrino paradigm and search for CP violation, requiring precise neutrino-flux normalization and improved control of accelerator-related uncertainties. Within the LBNF beamline, the System for On-Axis Neutrino Detection (SAND) will constrain flux uncertainties using precision near-detector measurements, while the Muon Monitor System (MuMS) will provide beamline diagnostics sensitive to the proton beam, target, and horn configuration. However, the pion phase space relevant for DUNE depends simultaneously on many correlated parameters, including beam centroid, beam width, horn current and alignment, target position, optics shifts, and radiation-induced changes. Consequently, MuMS observables exhibit nonlinear and coupled responses that are difficult to characterize using traditional one-parameter scans. To address this challenge, we are developing a Bayesian Exploration framework coupled to physics-informed surrogate emulators trained on Geant4 beamline simulations. Gaussian-process emulators provide both fast predictions and uncertainty estimates, enabling adaptive selection of new simulation points in beam-parameter space. As an initial demonstration, we construct surrogate emulators for MuMS response observables using a verified simulation campaign spanning proton-beam steering conditions. The emulators reproduce the simulated dependence of MuMS centroid and gradient observables while providing predictive uncertainties, and serve as the foundation for future multidimensional exploration including beam width, horn current, and additional beamline parameters. This work establishes a framework for uncertainty-aware beam monitoring, adaptive simulation campaigns, and rapid beam-response inference for future DUNE operations.

Ganguly, Sudeshna [Fermilab] (ORCID:00000003163482↗

DOE OSTI2025

Predicting Flow in Fracture Networks With Quantum Algorithms

Uncertainty quantification plays a crucial role in the modeling of subsurface flow. For instance, uncertainties in the properties of geologic fracture networks significantly impact flow, requiring numerous simulations to accurately estimate quantities of interest. However, each simulation is computationally expensive because it requires solving a large linear system to capture features that involve both small and large fractures. An example is in percolation, where the interaction of many small fractures (which cumulatively can have a large surface area) with the rock matrix must be modeled precisely. Quantum computing is an emerging tool with the potential to address this issue. Quantum algorithms offer a significant speedup in solving linear systems, achieving efficiencies that are challenging to match with classical approaches. These classical approaches include direct solvers, such as LU decomposition, and iterative methods, notably preconditioned conjugate gradient, commonly used in subsurface modeling to solve large sparse systems. However, applying quantum algorithms to geologic fracture flow requires careful attention to algorithmic and problem-specific constraints to fully realize this quantum advantage. In this work we describe a quantum algorithm for generalized Monte Carlo applications with a quadratic speedup over the classical approaches which can be combined with the quantum speedup, currently under investigation, for solving quantum linear systems for subsurface flow. We show that for quantum algorithms the computational cost of estimating a quantity of interest for a statistical ensemble of networks is roughly the same as that of a single realization, essentially implying that one can get uncertainty quantification for free.

58 GEOSCIENCES↗

DOE OSTI2026

Uncovering hidden bias in neutron diffraction residual strain measurements

When calculating residual strain via neutron or X-ray diffraction, uncertainties propagated from the peak fit are often inadequate to describe the true scatter of measurements about a singular strain state, such as one that should describe a macroscopic continuum. Because diffraction is inherently a selective process, orientation-dependent scatter arises from the sub-sampling of strong microstructure and strain gradients. This paper investigates the appropriateness of propagated uncertainties with reference to their original intention, i.e. noise about a mean value. Thirty-six unique orientations of strain measurements are taken at multiple locations within an additive friction-stir deposition component with fine-scale gradients (∼200 µm) of plastic strain, texture and residual elastic strain. Multiple strain and stress calculation pathways are compared: direct substitution of three measurements into Hooke's law, direct inversion of any six unique orientations into the strain state tensor and thirty-six measurement least-squares estimation. For the last two cases, the appropriateness of the uncertainty interval is statistically evaluated on the basis of a physical constraint: common agreement under the strain transformation law. For this sample, the direct inversion of six measurements retains a conservative estimate of the uncertainty. However, propagated uncertainties in the least-squares solution greatly underestimate the true experimental scatter. A simple pathway to estimate appropriate uncertainty intervals is suggested. These results demonstrate that the interpretation of uncertainty in residual strain is strongly dependent on intrinsic sample-dependent effects, and that oversampling orientations and statistical analysis can give more accurate results with realistic uncertainties.

36 MATERIALS SCIENCE↗

DOE OSTI2026

Cohort organized learning: clustering through agreement

In this article we describe cohort organized learning (CoOL), a method for clustering data without explicit distance or similarity computations. Herein, we will describe CoOL, derive the gradients determined by expectation maximization to train the networks, show how to monitor convergence during training and evaluate the clusters after training, and discuss a series of examples and use cases. We also discuss CoOL’s limitations and future prospects on related tasks. Because CoOL uses neural networks to estimate the clusters, it can be used to cluster any data that can be made compatible and we illustrate this on vector data and images.

clustering↗

DOE OSTI2026

Classical-Quantum Algorithm for Solving Stochastic Programs

Stochastic programming provides a rigorous mathematical framework for making decisions under uncertainty in a risk-aware manner. Two-stage stochastic programming is, perhaps, the simplest form of this framework. Here the first-stage variables represent decisions that must be made "here and now" in the face of uncertainty, while the second-stage variables are decisions made after uncertain events. However, the broad adoption of stochastic programming has been hindered by computational challenges caused by the two-stage stochastic programming formulation which requires solving an ensemble of optimization problems. Using quantum amplitude estimation (QAE), quantum computers have shown the theoretic ability to compute expectations with Monte-Carlo methods with quadratically fewer samples than classical methods. In this work, we present a quantum algorithm for computing the expectation term using QAE for given first-stage decisions. Further, we detail methods of computing gradient information from the quantum calculation enabling the application of classical gradient-based optimization techniques. The result is a classical-quantum hybrid method of solving two-stage stochastic programs. These techniques are demonstrated with computational experiments based an engineering optimization problem.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Neural Active Manifolds: Nonlinear Dimensionality Reduction for Uncertainty Quantification

We present a new approach for nonlinear dimensionality reduction, specifically designed for computationally expensive mathematical models. We leverage autoencoders to discover a one-dimensional neural active manifold (NeurAM) capturing the model output variability, through the aid of a simultaneously learnt surrogate model with inputs on this manifold. Our method only relies on model evaluations and does not require the knowledge of gradients. The proposed dimensionality reduction framework can then be applied to assist outer loop many-query tasks in scientific computing, like sensitivity analysis and multifidelity uncertainty propagation. In particular, we prove, both theoretically under idealized conditions, and numerically in challenging test cases, how NeurAM can be used to obtain multifidelity sampling estimators with reduced variance by sampling the models on the discovered low-dimensional and shared manifold among models. Several numerical examples illustrate the main features of the proposed dimensionality reduction strategy and highlight its advantages with respect to existing approaches in the literature.

Autoencoders↗