SEARCH · Search NASA

Results for “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 73 records · Page 4

All sources NASA NTRS DOE OSTI

DOE OSTI2022

Improving net ecosystem CO2 flux prediction using memory-based interpretable machine learning

Terrestrial ecosystems play a central role in the global carbon cycle and affect climate change. However, our predictive understanding of these systems is still limited due to their complexity and uncertainty about how key drivers and their legacy effects influence carbon fluxes. Here, we propose an interpretable Long Short-Term Memory (iLSTM) network for predicting net ecosystem CO 2 exchange (NEE) and interpreting the influence on the NEE prediction from environmental drivers and their memory effects. We consider five drivers and apply the method to three forest sites in the United States. Besides performing the prediction in each site, we also conduct transfer learning by using the iLSTM model trained in one site to predict at other sites. Results show that the iLSTM model produces good NEE predictions for all three sites and, more importantly, it provides reasonable interpretations on the input driver's importance as well as their temporal importance on the NEE prediction. Additionally, the iLSTM model demonstrates good across-site transferability in terms of both prediction accuracy and interpretability. The transferability can improve the NEE prediction in unobserved forest sites, and the interpretability advances our predictive understanding and guides process-based model development.

Liu, Siyan↗

DOE OSTI2015

MO-F-CAMPUS-I-01: Accuracy of Radiologists Interpretation of Mammographic Breast Density

Purpose: Several commercial and non-commercial software and techniques are available for determining breast density from mammograms. However, where mandated by law the breast density information communicated to the subject/patient is based on radiologist’s interpretation of breast density from mammograms. Several studies have reported on the concordance among radiologists in interpreting mammographic breast density. In this work, we investigated the accuracy of radiologist’s interpretation of breast density. Methods: Volumetric breast density (VBD) determined from 134 unilateral dedicated breast CT scans from 134 subjects was considered the truth. An MQSA-qualified study radiologist with more than 20 years of breast imaging experience reviewed the DICOM “for presentation” standard 2-view mammograms of the corresponding breasts and assigned BIRADS breast density categories. For statistical analysis, the breast density categories were dichotomized in two ways; fatty vs. dense breasts where “fatty” corresponds to BIRADS breast density categories A/B, and “dense” corresponds to BIRADS breast density categories C/D, and extremely dense vs. fatty to heterogeneously dense breasts, where extremely dense corresponds to BIRADS breast density category D and BIRADS breast density categories A through C were grouped as fatty to heterogeneously dense breasts. Logistic regression models (SAS 9.3) were used to determine the association between radiologist’s interpretation of breast density and VBD from breast CT, from which the area under the ROC (AUC) was determined. Results: Both logistic regression models were statistically significant (Likelihood Ratio test, p<0.0001). The accuracy (AUC) of the study radiologist for classification of fatty vs. dense breasts was 88.4% (95% CI: 83–94%) and for classification of extremely dense breast was 94.3% (95% CI: 90–98%). Conclusion: The accuracy of the radiologist in classifying dense and extremely dense breasts is high. Considering the variability in VBD estimates from commercial software, the breast density information communicated to the patient should be based on radiologist’s interpretation. This work was supported in part by NIH R21 CA176470 and R21 CA134128. The contents are solely the responsibility of the authors and do not reflect the official views of the NIH or NCI.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2023

Development of fracture diagnostic methods for fluid distribution based on quantitative interpretation of distributed acoustic sensing and distributed temperature sensing

Multistage hydraulic fracturing design on horizontal wells has significantly evolved with larger fluid volume, more fracturing stages, and tighter perforation cluster spacing to efficiently stimulate unconventional reservoirs. From the published field observations, the recent fracturing design results in complex fracture networks or swarm of fractures. Fracture treatment evaluation is extremely challenging in such a case because of the large amount of variables in well completion and stimulation design. Combined measurements from different technologies can help in fracture diagnosis. Fluid distribution, either during fracture injection or during production, directly relates to the stimulation efficiency at the cluster level and at the stage level. Because it is unlikely in the real world to distribute the injected fluid uniformly among all the clusters, we need diagnostic techniques to generate the flow profile along a lateral. Fiber-optic measurements, such as distributed acoustic sensing (DAS) and distributed temperature sensing (DTS), are currently used to diagnose downhole flow conditions. This technology allows us to qualitatively confirm the fluid flow profile and other issues occurring downhole during fracturing such as leakage through plugs. For optimizing a fracturing design, we also need to understand how the design parameters are correlated with the stimulation efficiency. In this study, we combine two sets of models of DAS and DTS data interpretation for injected fluid volume distribution. The DAS is interpreted based on an empirical correlation between fluid flow rates and frequency band energy from the acoustic signals. The DTS is interpreted by performing temperature history match-based thermal energy conservation. Because of the completely different physics behind the interpretations, the confirmation of two interpretations provides confidence in fluid distribution.

Geochemistry & Geophysics↗

DOE OSTI2018

Data-Driver Interpretation of Ultradeep Azimuthal Propagation

We report the interpretation of geological features around the borehole stresses on the search of true earth model parameters. Different from the conventional logging-while-drilling (LWD) technologies, such as wireline induction and propagation resistivity tools, the azimuthal resistivity measurement acquires the ability of directional sensitivity. Within this decade, this type of tools has been greatly enhanced with over hundred feet depth-of-detection (DoD). To resolve the interpretation of geological features within 100 feet from the borehole, the multiple-layer structure must be considered. However, the traditional methods can never handle the highly nonlinear inverse problems especially facing on the complicated earth structure. Meanwhile, no justifiable uncertainty of inverse parameters can be provided. The statistical methods have been adopted to search for the global statistical distribution and uncertainties of earth model parameters. However, the most attempts are still stuck in non-uniqueness of the solutions since the structure of earth model has been predefined, where the number of layers is assumed, which introduces another ambiguity into the model interpretation. This paper develops a general trans-dimensional Bayesian methodology to solve non-linear inverse problem of earth layer resistivity and depth to the layer boundaries, from the interpretation of ultra-deep azimuthal resistivity measurements. The presented work is a data-driven approach to search out the multi-dimensional parameter spaces. Finally, the simulations tell that the algorithm can efficiently solve the earth model interpretation within a hundred feet from borehole, and examine the uncertainty of earth model parameters.

58 GEOSCIENCES↗

DOE OSTI2021

Testing the reliability of interpretable neural networks in geoscience using the Madden–Julian oscillation

Abstract. We test the reliability of two neural network interpretation techniques, backward optimization and layerwise relevance propagation, within geoscientific applications by applying them to a commonly studied geophysical phenomenon, the Madden–Julian oscillation. The Madden–Julian oscillation is a multi-scale pattern within the tropical atmosphere that has been extensively studied over the past decades, which makes it an ideal test case to ensure the interpretability methods can recover the current state of knowledge regarding its spatial structure. The neural networks can, indeed, reproduce the current state of knowledge and can also provide new insights into the seasonality of the Madden–Julian oscillation and its relationships with atmospheric state variables. The neural network identifies the phase of the Madden–Julian oscillation twice as accurately as a linear regression approach, which means that nonlinearities used by the neural network are important to the structure of the Madden–Julian oscillation. Interpretations of the neural network show that it accurately captures the spatial structures of the Madden–Julian oscillation, suggest that the nonlinearities of the Madden–Julian oscillation are manifested through the uniqueness of each event, and offer physically meaningful insights into its relationship with atmospheric state variables. We also use the interpretations to identify the seasonality of the Madden–Julian oscillation and find that the conventionally defined extended seasons should be shifted later by 1 month. More generally, this study suggests that neural networks can be reliably interpreted for geoscientific applications and may thereby serve as a dependable method for testing geoscientific hypotheses.

58 GEOSCIENCES↗

DOE OSTI2021

The interpretation of temperature and salinity variables in numerical ocean model output and the calculation of heat fluxes and heat content

Abstract. The international Thermodynamic Equation of Seawater 2010 (TEOS-10) defined the enthalpy and entropy of seawater, thus enabling the global ocean heat content to be calculated as the volume integral of the product of in situ density, ρ, and potential enthalpy, h0 (with reference sea pressure of 0 dbar). In terms of Conservative Temperature, Θ, ocean heat content is the volume integral of ρcp0Θ, where cp0 is a constant “isobaric heat capacity”. However, many ocean models in the Coupled Model Intercomparison Project Phase 6 (CMIP6) as well as all models that contributed to earlier phases, such as CMIP5, CMIP3, CMIP2, and CMIP1, used EOS-80 (Equation of State – 1980) rather than the updated TEOS-10, so the question arises of how the salinity and temperature variables in these models should be physically interpreted, with a particular focus on comparison to TEOS-10-compliant observations. In this article we address how heat content, surface heat fluxes, and the meridional heat transport are best calculated using output from these models and how these quantities should be compared with those calculated from corresponding observations. We conclude that even though a model uses the EOS-80, which expects potential temperature as its input temperature, the most appropriate interpretation of the model's temperature variable is actually Conservative Temperature. This perhaps unexpected interpretation is needed to ensure that the air–sea heat flux that leaves and arrives in atmosphere and sea ice models is the same as that which arrives in and leaves the ocean model. We also show that the salinity variable carried by present TEOS-10-based models is Preformed Salinity, while the salinity variable of EOS-80-based models is also proportional to Preformed Salinity. These interpretations of the salinity and temperature variables in ocean models are an update on the comprehensive Griffies et al. (2016) paper that discusses the interpretation of many aspects of coupled Earth system models.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Multiscale graph neural network autoencoders for interpretable scientific machine learning

The goal of this work is to address two limitations in autoencoder-based models: latent space interpretability and compatibility with unstructured meshes. This is accomplished here with the development of a novel graph neural network (GNN) autoencoding architecture with demonstrations on complex fluid flow applications. To address the first goal of interpretability, the GNN autoencoder achieves reduction in the number nodes in the encoding stage through an adaptive graph reduction procedure. Further, this reduction procedure essentially amounts to flowfieldconditioned node sampling and sensor identification, and produces interpretable latent graph representations tailored to the flowfield reconstruction task in the form of so-called masked fields. These masked fields allow the user to (a) visualize where in physical space a given latent graph is active, and (b) interpret the time-evolution of the latent graph connectivity in accordance with the time-evolution of unsteady flow features (e.g. recirculation zones, shear layers) in the domain. To address the goal of unstructured mesh compatibility, the autoencoding architecture utilizes a series of multi-scale message passing (MMP) layers, each of which models information exchange among node neighborhoods at various lengthscales. The MMP layer, which augments standard single-scale message passing with learnable coarsening operations, allows the decoder to more efficiently reconstruct the flowfield from the identified regions in the masked fields. Analysis of latent graphs produced by the autoencoder for various model settings are conducted using unstructured snapshot data sourced from large-eddy simulations in a backward-facing step (BFS) flow configuration with an OpenFOAM-based flow solver at high Reynolds numbers.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

A detailed study of interpretability of deep neural network based top taggers

Abstract Recent developments in the methods of explainable artificial intelligence (XAI) allow researchers to explore the inner workings of deep neural networks (DNNs), revealing crucial information about input–output relationships and realizing how data connects with machine learning models. In this paper we explore interpretability of DNN models designed to identify jets coming from top quark decay in high energy proton–proton collisions at the Large Hadron Collider. We review a subset of existing top tagger models and explore different quantitative methods to identify which features play the most important roles in identifying the top jets. We also investigate how and why feature importance varies across different XAI metrics, how correlations among features impact their explainability, and how latent space representations encode information as well as correlate with physically meaningful quantities. Our studies uncover some major pitfalls of existing XAI methods and illustrate how they can be overcome to obtain consistent and meaningful interpretation of these models. We additionally illustrate the activity of hidden layers as neural activation pattern diagrams and demonstrate how they can be used to understand how DNNs relay information across the layers and how this understanding can help to make such models significantly simpler by allowing effective model reoptimization and hyperparameter tuning. These studies not only facilitate a methodological approach to interpreting models but also unveil new insights about what these models learn. Incorporating these observations into augmented model design, we propose the particle flow interaction network model and demonstrate how interpretability-inspired model augmentation can improve top tagging performance.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

SRBench++: Principled Benchmarking of Symbolic Regression With Domain-Expert Interpretation

Symbolic regression searches for analytic expressions that accurately describe studied phenomena. The main promise of this approach is that it may return an interpretable model that can be insightful to users, while maintaining high accuracy. The current standard for benchmarking these algorithms is SRBench, which evaluates methods on hundreds of datasets that are a mix of real-world and simulated processes spanning multiple domains. At present, the ability of SRBench to evaluate interpretability is limited to measuring the size of expressions on real-world data, and the exactness of model forms on synthetic data. In practice, model size is only one of many factors used by subject experts to determine how interpretable a model truly is. Furthermore, SRBench does not characterize algorithm performance on specific, challenging sub-tasks of regression such as feature selection and evasion of local minima. In this work, we propose and evaluate an approach to benchmarking SR algorithms that addresses these limitations of SRBench by 1) incorporating expert evaluations of interpretability on a domain-specific task, and 2) evaluating algorithms over distinct properties of data science tasks. We evaluate 12 modern symbolic regression algorithms on these benchmarks and present an in-depth analysis of the results, discuss current challenges of symbolic regression algorithms and highlight possible improvements for the benchmark itself.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Higgs Portal Vector Dark Matter Interpretation: Review of Effective Field Theory Approach and Ultraviolet Complete Models

In this report a review of the Higgs portal-vector dark matter interpretation of the spin-independent dark matter nu-cleon elastic scattering cross section is presented, where the invisible Higgs decay width measured at the LHC is used. Effective Field Theory and ultraviolet complete models are discussed. LHC interpretations show only the scalar and Majorana dark matter scenarios; we propose including interpretation for vec-tor dark matter in the EFT and UV completion theoretical framework. In addition, our studies suggest an extension of the LHC dark matter interpretations to the sub-GeV regime.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2019

Orbitals and the Interpretation of Photoelectron Spectroscopy and (e,2e) Ionization Experiments

Abstract Electron momentum spectroscopy, scanning tunneling microscopy, and photoelectron spectroscopy provide unique information about electronic structure, but their interpretation has been controversial. This essay discusses a framework for interpretation. Although this interpretation is not new, we believe it is important to present this framework in light of recent publications. The key point is that these experiments provide information about how the electron distribution changes upon ionization, not how electrons behave in the pre‐ionized state. Therefore, these experiments do not lead to a “selection of the correct orbitals” in chemistry and do not overturn the well‐known conclusion that both delocalized molecular orbitals and localized molecular orbitals are useful for interpreting chemical structure and dynamics. The two types of orbitals can produce identical total molecular electron densities and therefore molecular properties. Different types of orbitals are useful for different purposes.

Truhlar, Donald G.↗

DOE OSTI2019

Orbitals and the Interpretation of Photoelectron Spectroscopy and (e,2e) Ionization Experiments

Truhlar, Donald G.↗

DOE OSTI2020

False positive PSMA PET for tumor remnants in the irradiated prostate and other interpretation pitfalls in a prospective multi-center trial

Readers need to be informed about potential pitfalls of [ 68 Ga]Ga-PSMA-11 PET interpretation. Here we report [ 68 Ga]Ga-PSMA-11 PET findings discordant with the histopathology/composite reference standard in a recently published prospective trial on 635 patients with biochemically recurrent prostate cancer. Consensus reads were false positive in 20 regions of 17/217 (8%) patients with lesion validation. Majority of the false positive interpretations (13 of 20, 65%) occurred in the context of suspected prostate (bed) relapse (T) after radiotherapy ( n = 11); other false positive findings were noted for prostate bed post prostatectomy (T, n = 2), pelvic nodes (N, n = 2), or extra pelvic lesions (M, n = 5). Major sources of false positive findings were PSMA-expressing residual adenocarcinoma with marked post-radiotherapy treatment effect. False negative interpretation occurred in 8 regions of 6/79 (8%) patients with histopathology validation, including prostate (bed) ( n = 5), pelvic nodes ( n = 1), and extra pelvic lesions ( n = 2). Lesions were missed mostly due to small metastases or adjacent bladder/urine uptake.[ 68 Ga]Ga-PSMA-11 PET at biochemical recurrence resulted in less than 10% false positive interpretations. Post-radiotherapy prostate uptake was a major source of [ 68 Ga]Ga-PSMA-11 PET false positivity. In few cases, PET correctly detects residual PSMA expression post-radiotherapy, originating however from treated, benign tissue or potentially indolent tumor remnants.

Radiology, Nuclear Medicine & Medical Imaging↗

DOE OSTI2018

An Automated, High-Throughput Method for Interpreting the Tandem Mass Spectra of Glycosaminoglycans

The biological interactions between glycosaminoglycans (GAGs) and other biomolecules are heavily influenced by structural features of the glycan. The structure of GAGs can be assigned using tandem mass spectrometry (MS{sup 2}), but analysis of these data, to date, requires manually interpretation, a slow process that presents a bottleneck to the broader deployment of this approach to solving biologically relevant problems. Automated interpretation remains a challenge, as GAG biosynthesis is not template-driven, and therefore, one cannot predict structures from genomic data, as is done with proteins. The lack of a structure database, a consequence of the non-template biosynthesis, requires a de novo approach to interpretation of the mass spectral data. We propose a model for rapid, high-throughput GAG analysis by using an approach in which candidate structures are scored for the likelihood that they would produce the features observed in the mass spectrum. To make this approach tractable, a genetic algorithm is used to greatly reduce the search-space of isomeric structures that are considered. The time required for analysis is significantly reduced compared to an approach in which every possible isomer is considered and scored. The model is coded in a software package using the MATLAB environment. This approach was tested on tandem mass spectrometry data for long-chain, moderately sulfated chondroitin sulfate oligomers that were derived from the proteoglycan bikunin. The bikunin data was previously interpreted manually. Our approach examines glycosidic fragments to localize SO{sub 3} modifications to specific residues and yields the same structures reported in literature, only much more quickly. .

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2023

Machine learning and deep learning for mineralogy interpretation and CO 2 saturation estimation in geological carbon Storage: A case study in the Illinois Basin

Carbon capture and storage (CCS) is a promising approach to simultaneously maintaining energy security and reducing carbon dioxide (CO 2 ) emissions under the current energy portfolio that is dominated by fossil fuel energy. Pre-injection formation characterization and post-injection CO 2 monitoring are two critical tasks to guarantee storage efficiency in CCS. The CCS projects in the Illinois Basin, the first large-scale CO 2 injection into saline aquifers in the United States, employed conventional and the latest pulsed neutron logging (PNL) tools for mineralogy interpretation and CO 2 saturation estimation, which provide valuable references for future CCS projects. Because of the inherent fuzziness of petrophysical measurements and complex subsurface heterogeneity, interpreting well-logging data is time-consuming, and its accuracy can be user-biased. In recent years, data-driven methods have been widely used to capture the non-linear patterns between input features and interpretation results. This work applied and evaluated four commonly used machine learning (ML) models, including ridge regression (RR), random forest (RF), gradient boosting regression (GBR), support vector regression (SVR), and one deep learning (DL) model, the artificial neural network (ANN). We optimized the hyperparameters of the four ML models and the DL model using the simulated annealing algorithm and the grid search strategy, respectively. The input features of the mineralogy interpretation models were eleven conventional well-logging parameters, and the label data (i.e., ground truth) were the porosity and volumetric fractions of six minerals, including quartz, feldspar, dolomite, calcite, clay, and iron minerals. The results demonstrated that the GBR and RF models were superior in predicting volumetric fractions of minerals and porosity; label data with low coefficient of variation (CV) values tended to yield better performance. For CO 2 saturation estimation, the RF was the best-performing model, followed by SVR, ANN, GBR, and RR. Furthermore, we conducted feature importance ranking using the permutation importance algorithm and found that the formation sigma and well pressure were the most important features in this study. In conclusion, the study of CCS projects in the Illinois Basin bridges the gap between the limited knowledge and understanding of geological carbon storage and the increasing demand for reliable, cost-effective, and sustainable energy solutions.

58 GEOSCIENCES↗

DOE OSTI2019

Validating compositional fluid flow simulations using 4D seismic interpretation and vice versa in the SECARB Early Test—A critical review

We report on strengths and weaknesses of 4D seismic interpretation as a technique for monitoring carbon dioxide that was injected as part of a large scale test associated with commercial enhanced oil recovery (EOR) at Cranfield Field, Mississippi, USA. The goals of the monitoring effort are 1) to make measurements to verify that the CO 2 is contained in the reservoir according to operational designs and model predictions, and 2) that if there are deviations, to provide data which can be used to update the earth models and determine if any mitigation is needed. Our work uses a compositional numerical simulation to model CO 2 flow in the reservoir and compares the results with estimates of changes in seismic properties between a pre-injection and a survey after more than 2 million metric tons injected. The complicated physics of the problem in Cranfield field present challenges to seismic interpretations. These findings reflect partial agreement between the results of the numerical simulation and the seismic interpretations. Possible causes of discrepancies among the fluid flow model and multiple interpretations of the same seismic data sets performed with different work flows illuminates the types of uncertainties that should be considered to achieve the goals of monitoring.

4D seismic↗

DOE OSTI2015

Interpretation of the photoelectron, ultraviolet, and vacuum ultraviolet photoabsorption spectra of bromobenzene by ab initio configuration interaction and DFT computations

New photoelectron, ultraviolet (UV), and vacuum UV (VUV) spectra have been obtained for bromobenzene by synchrotron study with higher sensitivity and resolution than previous work. This, together with use of ab initio calculations with both configuration interaction and time dependent density functional theoretical methods, has led to major advances in interpretation. The VUV spectrum has led to identification of a considerable number of Rydberg states for the first time. The Franck-Condon (FC) analyses including both hot and cold bands lead to identification of the vibrational structure of both ionic and electronically excited states including two Rydberg states. The UV onset has been interpreted in some detail, and an interpretation based on the superposition of FC and Herzberg-Teller contributions has been performed. In a similar way, the 6 eV absorption band which is poorly resolved is analysed in terms of the presence of two ππ* states of {sup 1}A{sub 1} (higher oscillator strength) and {sup 1}B{sub 2} (lower oscillator strength) symmetries, respectively. The detailed analysis of the vibrational structure of the 2{sup 2}B{sub 1} ionic state is particularly challenging, and the best interpretation is based on equation-of-motion-coupled cluster with singles and doubles computations. A number of equilibrium structures of the ionic and singlet excited states show that the molecular structure is less subject to variation than corresponding studies for iodobenzene. The equilibrium structures of the 3b{sub 1}3s and 6b{sub 2}3s (valence shell numbering) Rydberg states have been obtained and compared with the corresponding ionic limit structures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2020

Uncovering interpretable relationships in high-dimensional scientific data through function preserving projections

Abstract In many fields of science and engineering, we frequently encounter experiments or simulations datasets that describe the behavior of complex systems and uncovering human interpretable patterns between their inputs and outputs via exploratory data analysis is essential for building intuition and facilitating discovery. Often, we resort to 2D embeddings for examining these high-dimensional relationships (e.g. dimensionality reduction). However, most existing embedding methods treat the dimensions as coordinates for samples in a high-dimensional space, which fail to capture the potential functional relationships, and the few methods that do take function into consideration either only focus on linear patterns or produce non-linear embeddings that are hard to interpret. To address these challenges, we proposed function preserving projections (FPP), which construct 2D linear embeddings optimized to reveal interpretable yet potentially non-linear patterns between the domain and the range of a high-dimensional function. The intuition here is that humans are good at understanding potentially non-linear patterns in 2D but unable to interpret non-linear mapping from high-dimensional space to 2D. Therefore, we should restrict the projection to linear but not the pattern we are seeking. Using FPP on real-world datasets, one can obtain fundamentally new insights about high-dimensional relationships in extremely large datasets that could not be processed with existing dimension reduction methods.

97 MATHEMATICS AND COMPUTING↗