SEARCH · Engineering Papers

Results for “model interpretability”

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 325 records · Page 18

All sources NASA NTRS DOE OSTI

NASA NTRS1990

Predicted distribution of visible and near-infrared radiant flux above and below a transmittant leaf

The effects are studied analytically of leaf size, leaf height, and background reflectance on the upward and downward radiant flux (RF) of a leaf. The leaf is horizontal and isotropically scattering in the computer model which examines the light environment in three regions about the leaf. The spectral properties of the leaf are based on measurements of the big-leaf maple, and the model is interpreted in terms of relative RF which is defined as a percentage of the total light in the model. The results demonstrate the dependence of upward relative RF on the light's wavelength and background reflectance with large variations in the NIR. Brightness varies directly with distance from background with maximum brightness achieved at lower heights for smaller leaves. These and other results suggest that NIR canopy reflectance due to leaves is highly dependent on the background reflectance.

Roberts, Dar A.↗

NASA NTRS1995

A relational metric, its application to domain analysis, and an example analysis and model of a remote sensing domain

An objective and quantitative method has been developed for deriving models of complex and specialized spheres of activity (domains) from domain-generated verbal data. The method was developed for analysis of interview transcripts, incident reports, and other text documents whose original source is people who are knowledgeable about, and participate in, the domain in question. To test the method, it is applied here to a report describing a remote sensing project within the scope of the Earth Observing System (EOS). The method has the potential to improve the designs of domain-related computer systems and software by quickly providing developers with explicit and objective models of the domain in a form which is useful for design. Results of the analysis include a network model of the domain, and an object-oriented relational analysis report which describes the nodes and relationships in the network model. Other products include a database of relationships in the domain, and an interactive concordance. The analysis method utilizes a newly developed relational metric, a proximity-weighted frequency of co-occurrence. The metric is applied to relations between the most frequently occurring terms (words or multiword entities) in the domain text, and the terms found within the contexts of these terms. Contextual scope is selectable. Because of the discriminating power of the metric, data reduction from the association matrix to the network is simple. In addition to their value for design. the models produced by the method are also useful for understanding the domains themselves. They can, for example, be interpreted as models of presence in the domain.

Mcgreevy, Michael W.↗

NASA NTRS2024

Interpretable Machine Learning for Molecular Biosignatures: a Novel Single-Sample Feature Importance Method That Is Sensitive To Statistical Interactions

Isotope ratio mass spectrometry (IRMS) of volatiles (e.g., CO 2 ) promises to be a powerful tool for potential biosignature detection for future missions to ocean worlds (OW) such as Europa and Enceladus. Machine learning (ML) methods for IRMS data could enable science autonomy by onboard prediction of seawater chemistry and biosignature presence. However, ML models are likely to be complex and involve statistical interactions between features (variables), which can make predictions seem opaque and enigmatic. For ML predictions as significant as extraterrestrial biosignatures, we must place extraordinary confidence in models. It is therefore essential that these models make interpretable predictions (i.e., human-understandable) and include false-prediction diagnostics. We achieve high accuracy and interpretability in ML biosignature and seawater chemistry models for OW through a nearest-neighbors feature selection tool that detects statistical interactions between predictors, constructs interaction networks for visualization of selected features working together to make a prediction, and reports single-sample feature importance scores for false-detection diagnostics. Here we develop a novel single-sample nearest-neighbors projected distance regression(ssNPDR) feature selection method that improves upon existing single-sample algorithms through the inclusion of statistical interactions while providing false-prediction diagnostics for ML models.

geochemistry↗

DOE OSTI2026

Learning thermodynamic master equations for open quantum systems

The characterization of Hamiltonians and other components of open quantum dynamical systems plays a crucial role in quantum computing and other applications. Scientific machine learning techniques have been applied to this problem in a variety of ways, including by modeling with deep neural networks. However, the majority of mathematical models describing open quantum systems are linear, and the natural nonlinearities in learnable models have not been incorporated using physical principles. We present a data-driven model for open quantum systems that includes learnable, thermodynamically consistent terms. The trained model is interpretable, as it directly estimates the system Hamiltonian and linear components of coupling to the environment. We validate the model on synthetic two and three-level data, as well as experimental two-level data collected from a quantum device at Lawrence Livermore National Laboratory.

Mathematics and Computing↗

NASA NTRS2015

Introduction to Adjoint Models

In this lecture, some fundamentals of adjoint models will be described. This includes a basic derivation of tangent linear and corresponding adjoint models from a parent nonlinear model, the interpretation of adjoint-derived sensitivity fields, a description of methods of automatic differentiation, and the use of adjoint models to solve various optimization problems, including singular vectors. Concluding remarks will attempt to correct common misconceptions about adjoint models and their utilization.

EnKF↗

DOE OSTI2026

Artificial Intelligence-Assisted Daytime Video Monitoring for Bird, Insect, and Other Wildlife Interactions with Photovoltaic Solar Energy Facilities

Studying bird, insect, and other wildlife interactions with photovoltaic (PV) solar energy facilities is difficult due to limited multi-season, multi-site data. Researchers can address such data gaps by combining passive monitoring and artificial intelligence (AI). As a part of the development of AI-enabled avian–solar monitoring software, we collected over 19,000 h of daytime videos at five PV sites across three U.S. regions between 2019 and 2024. We applied a moving object detection and tracking (MODT Version 1) AI model we developed earlier to 4373 h of the footage to extract moving objects in video frames, and human reviewers interpreted the model output and identified 68,646 bird, 25,968 insect, and 169 other wildlife instances to generate the training/validation dataset. We analyzed the data by site, region, and season, considering ground cover and landscapes. Songbirds were most common, with raptors as the next most frequent group. Most notably, no bird collisions were confirmed in our observations collected from the videos. Birds most often flew over or near panels, with the highest observations in the Midwest and Northeast (approximately 30 observations per hour on average) and fewer in the desert Southwest. Other behaviors included perching, foraging, and nesting. Bird abundance peaked during breeding and migration seasons. AI-assisted video monitoring proved effective for non-invasively studying flying wildlife at solar facilities to inform ecologically mindful energy development.

avian mortality↗

NASA NTRS1994

Empirical convection models for northward IMF

It is clear that polar cap convection during times of northward Interplanetary Magnetic Field (IMF) is more structured and of lower mean speed than at times of southward IMF. This, coupled with the fact that the polar cap is smaller, means that empirical models are more difficult to construct with certainty. It is also clear that sunward flow deep in the polar cap is often observed, but its connection with the rest of the flow pattern is controversial. At present, empirical models are of three types: 'statistical' models wherein data from different days but with similar IMF conditions are averaged together; 'pattern recognition' models, which are built up by examining individually hundreds of passes to derive a 'typical' pattern which embodies features frequently observed; and 'assimilative' models, which use data of different types and from as many locations as possible, but all taken at the same time, in order to derive a snapshot (or series of snapshots) of the entire pattern. Each type of model has its own difficulties. Statistical models, by their very nature, smooth out flow features (e.g. the convection reversal, and the locus of sunward flow deep in the polar cap) which are not found at precisely the same invariant latitudes and magnetic local times on different days. Pattern recognition models are better at reproducing small-scale features, but the large-scale pattern can be a matter of interpretation. Assimilative models (such as AMIE) hold out the best hope for creating instantaneous, global convection patterns; however, the analysis technique tends to be most irregular (and least reliable) in the regions which are not well covered by in situ data. It appears that, at least at times, a four cell model with sunward flow at the highest and lowest latitudes, and antisunward flow in between, is consistent with the observations. At other times, the observations may be consistent with a two-cell convection pattern, but which includes significant meanders within the polar cap.

Moses, Julie J.↗

NASA NTRS1989

Multiresolutional models of uncertainty generation and reduction

Kolmogorov's axiomatic principles of the probability theory, are reconsidered in the scope of their applicability to the processes of knowledge acquisition and interpretation. The model of uncertainty generation is modified in order to reflect the reality of engineering problems, particularly in the area of intelligent control. This model implies algorithms of learning which are organized in three groups which reflect the degree of conceptualization of the knowledge the system is dealing with. It is essential that these algorithms are motivated by and consistent with the multiresolutional model of knowledge representation which is reflected in the structure of models and the algorithms of learning.

Meystel, A.↗

NASA NTRS1982

Age and provenance of the target materials for tektites and possible impactites as inferred from Sm-Nd and Rb-Sr systematics

Chemical, trace element, and isotopic compositions of tektites are consistent with production by melting of sediments derived from the old terrestrial continental crust. Each tektite group is characterized by a uniform Nd model age, interpreted as the time of formation of the crustal segment which weathered to form the parent sediment for the tektites. Sr model ages are variable within each group, reflecting Rb-Sr fractionation, and, in the favorable limit of very high Rb/Sr ratios, approach the time of sedimentation of the parent material which melted to form the tektites. Unlike tektites, which are dense homogeneous objects, sanidine spherules are porous, fine grained inhomogeneous objects. The leaching experiment employed by the present study shows that the sanidine spherules could have been formed by an oceanic impact involving basaltic crust and overlying sediments or seawater.

Shaw, H. F.↗

NASA NTRS1995

Ground-based grasslands data to support remote sensing and ecosystem modeling of terrestrial primary production

Estimating terrestrial net primary production (NPP) using remote-sensing tools and ecosystem models requires adequate ground-based measurements for calibration, parameterization, and validation. These data needs were strongly endorsed at a recent meeting of ecosystem modelers organized by the International Geosphere-Biosphere Program's (IGBP's) Data and Information System (DIS) and its Global Analysis, Interpretation, and Modelling (GAIM) Task Force. To meet these needs, a multinational, multiagency project is being coordinated by the IGBP DIS to compile existing NPP data from field sites and to regionalize NPP point estimates to various-sized grid cells. Progress at Oak Ridge National Laboratory (ORNL) on compiling NPP data for grasslands as part of the IGBP DIS data initiative is described. Site data and associated documentation from diverse field studies are being acquired for selected grasslands and are being reviewed for completeness, consistency, and adequacy of documentation, including a description of sampling methods. Data are being compiled in a database with spatial, temporal, and thematic characteristics relevant to remote sensing and global modeling. NPP data are available from the ORNL Distributed Active Archive Center (DAAC) for biogeochemical dynamics. The ORNL DAAC is part of the Earth Observing System Data and Information System, of the US National Aeronautics and Space Administration.

Olson, R. J.↗

DOE OSTI2025

A hybrid calorimetry-simulation model of mixing enthalpy for molten salt

Calorimetric determination of enthalpies of mixing (ΔH mix ) in multicomponent molten salts is often interpreted using empirical models that lack physically meaningful parameters. However, for improving pyrochemical separation of spent nuclear fuel, where lanthanides are major fission products and critical elements, a deeper thermodynamic understanding of the link between excess thermodynamic properties and solvation structure is critically needed. In this work, we implement a hybrid and physics-informed framework, MIVM+Calorimetry+AIMD, which integrates experimentally measured ΔH mix (via high temperature drop calorimetry) with solvation structures from ab initio molecular dynamics (AIMD). This approach is demonstrated using LaCl 3 mixed with eutectic LiCl-KCl (58 mol% – 42 mol%) at 873 K and 1133 K. MIVM-derived parameters enable extrapolation of excess Gibbs energy and La 3+ activity across compositions. In contrast, direct ΔH mix predictions from AIMD and polarizable ion model simulations deviate significantly. By incorporating experimentally benchmarked solvation structures into an interpretable thermodynamic model, the MIVM+Calorimetry+AIMD formalism achieves higher accuracy and generalizable method for studying molten salts, offering a robust path for understanding and optimizing molten salt chemistry relevant to nuclear fuel cycles and separation science.

Goncharov, Vitaliy G. [Washington State Univ., Pul↗

NASA NTRS2000

Sensitivity of Age-of-Air Calculations to the Choice of Advection Scheme

The age of air has recently emerged as a diagnostic of atmospheric transport unaffected by chemical parameterizations, and the features in the age distributions computed in models have been interpreted in terms of the models' large-scale circulation field. This study shows, however, that in addition to the simulated large-scale circulation, three-dimensional age calculations can also be affected by the choice of advection scheme employed in solving the tracer continuity equation, Specifically, using the 3.0deg latitude X 3.6deg longitude and 40 vertical level version of the Geophysical Fluid Dynamics Laboratory SKYHI GCM and six online transport schemes ranging from Eulerian through semi-Lagrangian to fully Lagrangian, it will be demonstrated that the oldest ages are obtained using the nondiffusive centered-difference schemes while the youngest ages are computed with a semi-Lagrangian transport (SLT) scheme. The centered- difference schemes are capable of producing ages older than 10 years in the mesosphere, thus eliminating the "young bias" found in previous age-of-air calculations. At this stage, only limited intuitive explanations can be advanced for this sensitivity of age-of-air calculations to the choice of advection scheme, In particular, age distributions computed online with the National Center for Atmospheric Research Community Climate Model (MACCM3) using different varieties of the SLT scheme are substantially older than the SKYHI SLT distribution. The different varieties, including a noninterpolating-in-the-vertical version (which is essentially centered-difference in the vertical), also produce a narrower range of age distributions than the suite of advection schemes employed in the SKYHI model. While additional MACCM3 experiments with a wider range of schemes would be necessary to provide more definitive insights, the older and less variable MACCM3 age distributions can plausibly be interpreted as being due to the semi-implicit semi-Lagrangian dynamics employed in the MACCM3. This type of dynamical core (employed with a 60-min time step) is likely to reduce SLT's interpolation errors that are compounded by the short-term variability characteristic of the explicit centered-difference dynamics employed in the SKYHI model (time step of 3 min). In the extreme case of a very slowly varying circulation, the choice of advection scheme has no effect on two-dimensional (latitude-height) age-of-air calculations, owing to the smooth nature of the transport circulation in 2D models. These results suggest that nondiffusive schemes may be the preferred choice for multiyear simulations of tracers not overly sensitive to the requirement of monotonicity (this category includes many greenhouse gases). At the same time, age-of-air calculations offer a simple quantitative diagnostic of a scheme's long-term diffusive properties and may help in the evaluation of dynamical cores in multiyear integrations. On the other hand, the sensitivity of the computed ages to the model numerics calls for caution in using age of air as a diagnostic of a GCM's large-scale circulation field.

Eluszkiewicz, Janusz↗

DOE OSTI2022

A Gauss’s law analysis of redox active adsorbates on semiconductor electrodes: The charging and faradaic currents are not independent

A detailed framework for modeling and interpreting the data in totality from a cyclic voltammetric measurement of adsorbed redox monolayers on semiconductor electrodes has been developed. A three-layer model consisting of the semiconductor space-charge layer, a surface layer, and an electrolyte layer is presented that articulates the interplay between electrostatic, thermodynamic, and kinetic factors in the electrochemistry of a redox adsorbate on a semiconductor. Expressions are derived that describe the charging and faradaic current densities individually, and an algorithm is demonstrated that allows for the calculation of the total current density in a cyclic voltammetry measurement as a function of changes in the physical properties of the system (e.g., surface recombination, dielectric property of the surface layer, and electrolyte concentration). The most profound point from this analysis is that the faradaic and charging current densities can be coupled. That is, the common assumption that these contributions to the total current are always independent is not accurate. Their interrelation can influence the interpretation of the charge-transfer kinetics under certain experimental conditions. More generally, this work not only fills a long-standing knowledge gap in electrochemistry but also aids practitioners advancing energy conversion/storage strategies based on redox adsorbates on semiconductor electrodes.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NASA NTRS2012

The Performance of Current Atmospheric Radiation Codes in Phase I of CIRC

The Continual Intercomparison of Radiation Codes (CIRC) is intended as an evolving and regularly updated reference source for evaluation of radiative transfer (RT) codes used in Global Climate Models and other atmospheric applications. In our presentation we will discuss our evaluation of the performance of 13 shortwave and 11 longwave RT codes that participated in Phase I of CIRC. CIRC differs from previous intercomparisons in that it relies on an observationally validated catalogue of cases. The seven CIRC Phase I baseline cases, five cloud-free, and two with overcast liquid clouds, are built around observations by the Atmospheric Radiation Measurements (ARM) program that satisfy the goals .of Phase I, namely to examine RT model performance in realistic, yet not overly complex, atmospheric conditions. Besides the seven baseline cases, additional idealized "subcases" are also examined to facilitate interpretation of model errors. We will quantify individual model performance with respect to reference line-by-line calculations, and will also highlight RT code behavior for conditions of doubled CO2 , aspects of utilizing a spectral specification of surface albedo, and the impact of the inclusion of scattering in the thermal infrared. Our analysis suggests that RT codes should work towards improving their calculation of diffuse shortwave flux, shortwave absorption, treatment of spectral surface albedo, and shortwave CO2 forcing. Despite practical difficulties in comparing our results to previous results by the Intercomparison of Radiation Codes in Climate Models (ICRCCM) conducted about 20 years ago, it appears that the current generation of RT codes do indeed perform better than the codes of the ICRCCM era. By enhancing the range of conditions under which participating codes are tested, future CIRC phases will hopefully allow even more rigorous examination of RT code performance.

Oreopoulos, L.↗

DOE OSTI2023

Differentiable modelling to unify machine learning and physical models for geosciences

Process-based modelling offers interpretability and physical consistency in many domains of geosciences but struggles to leverage large datasets efficiently. Machine-learning methods, especially deep networks, have strong predictive skills yet are unable to answer specific scientific questions. Here, in this Perspective, we explore differentiable modelling as a pathway to dissolve the perceived barrier between process-based modelling and machine learning in the geosciences and demonstrate its potential with examples from hydrological modelling. ‘Differentiable’ refers to accurately and efficiently calculating gradients with respect to model variables or parameters, enabling the discovery of high-dimensional unknown relationships. Differentiable modelling involves connecting (flexible amounts of) prior physical knowledge to neural networks, pushing the boundary of physics-informed machine learning. It offers better interpretability, generalizability, and extrapolation capabilities than purely data-driven machine learning, achieving a similar level of accuracy while requiring less training data. Additionally, the performance and efficiency of differentiable models scale well with increasing data volumes. Under data-scarce scenarios, differentiable models have outperformed machine-learning models in producing short-term dynamics and decadal-scale trends owing to the imposed physical constraints. Differentiable modelling approaches are primed to enable geoscientists to ask questions, test hypotheses, and discover unrecognized physical relationships. Future work should address computational challenges, reduce uncertainty, and verify the physical significance of outputs.

58 GEOSCIENCES↗

DOE OSTI2023

A search for new resonances in multiple final states with a high transverse momentum Z boson in $\sqrt{s} $ = 13 TeV pp collisions with the ATLAS detector

A generic search for resonances is performed with events containing a Z boson with transverse momentum greater than 100 GeV, decaying into e + e – or μ + μ – . The analysed data collected with the ATLAS detector in proton-proton collisions at a centre-of-mass energy of 13 TeV at the Large Hadron Collider correspond to an integrated luminosity of 139 fb –1 . Two invariant mass distributions are examined for a localised excess relative to the expected Standard Model background in six independent event categories (and their inclusive sum) to increase the sensitivity. No significant excess is observed. Exclusion limits at 95% confidence level are derived for two cases: a model-independent interpretation of Gaussian-shaped resonances with the mass width between 3% and 10% of the resonance mass, and a specific heavy vector triplet model with the decay mode W' → ZW → ℓℓqq.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2023

maslov-group/FUN-PROSE

mRNA levels of all genes in a genome is a critical piece of information defining the overall state of the cell in a given environmental condition. Being able to reconstruct such condition-specific expression in fungal genomes is particularly important to metabolically engineer these organisms to produce desired chemicals in industrially scalable conditions. Most previous deep learning approaches focused on predicting the average expression levels of a gene based on its promoter sequence, ignoring its variation across different conditions. Here we present FUN-PROSE—a deep learning model trained to predict differential expression of individual genes across various conditions using their promoter sequences and expression levels of all transcription factors. We train and test our model on three fungal species and get the correlation between predicted and observed condition-specific gene expression as high as 0.85. We then interpret our model to extract promoter sequence motifs responsible for variable expression of individual genes. We also carried out input feature importance analysis to connect individual transcription factors to their gene targets. A sizeable fraction of both sequence motifs and TF-gene interactions learned by our model agree with previously known biological information, while the rest corresponds to either novel biological facts or indirect correlations.

Liu, Simon↗

DOE OSTI2023

FUN-PROSE: A deep learning approach to predict condition-specific gene expression in fungi

59 BASIC BIOLOGICAL SCIENCES↗