Extracting an Empirical Intermetallic Hydride Design Principle from Limited Data via Interpretable Machine Learning.
Abstract not provided.
Witman, Matthew↗
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.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Calculation and interpretation of classical turning surfaces in solids
Abstract Classical turning surfaces of Kohn–Sham potentials separate classically allowed regions (CARs) from classically forbidden regions (CFRs). They are useful for understanding many chemical properties of molecules but need not exist in solids, where the density never decays to zero. At equilibrium geometries, we find that CFRs are absent in perfect metals, rare in covalent semiconductors at equilibrium, but common in ionic and molecular crystals. In all materials, CFRs appear or grow as the internuclear distances are uniformly expanded. They can also appear at a monovacancy in a metal. Calculations with several approximate density functionals and codes confirm these behaviors. A classical picture of conduction suggests that CARs should be connected in metals, and disconnected in wide-gap insulators, and is confirmed in the limits of extreme compression and expansion. Surprisingly, many semiconductors have no CFR at equilibrium, a key finding for density functional construction. Nonetheless, a strong correlation with insulating behavior can still be inferred. Moreover, equilibrium bond lengths for all cases can be estimated from the bond type and the sum of the classical turning radii of the free atoms or ions.
Chemistry↗
HAWC observations strongly favor pulsar interpretations of the cosmic-ray positron excess
Not Available
Hooper, Dan↗
Interpretation of proton radiography experiments of hohlraums with three-dimensional simulations [Interpretation of proton radiography experiments of hohlraums with 3D simulations]
Proton radiography experiments of laser-irradiated hohlraums performed at the OMEGA laser facility are analyzed using three-dimensional (3D) hydrodynamic simulations coupled to a proton trajectography package. Experiments with three different laser irradiation patterns were performed, and each produced a distinct proton image. By comparing these results with synthetic proton images obtained by sending protons through plasma profiles in the hohlraum obtained from 3D radiation hydrodynamic simulations, it is found that the simulated images agree favorably with the experimental images when electric fields, due to the electron pressure gradients that arise from 3D structures occurring during plasma expansion, are included. Furthermore, these comparisons provide quantitative estimates of the electric field present inside the hohlraums.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Time to Go Beyond Triple-Gauge-Boson-Coupling Interpretation of W Pair Production
Not Available
Zhang, Zhengkang↗
Detection of regional infrasound signals using array data: Testing, tuning, and physical interpretation
This work quantifies the physical characteristics of infrasound signal and noise, assesses their temporal variations, and determines the degree to which these effects can be predicted by time-varying atmospheric models to estimate array and network performance. An automated detector that accounts for both correlated and uncorrelated noise is applied to infrasound data from three seismo-acoustic arrays in South Korea (BRDAR, CHNAR, and KSGAR), cooperatively operated by Korea Institute of Geoscience and Mineral Resources (KIGAM) and Southern Methodist University (SMU). Arrays located on an island and near the coast have higher noise power, consistent with both higher wind speeds and seasonably variable ocean wave contributions. On the basis of the adaptive F-detector quantification of time variable environmental effects, the time-dependent scaling variable is shown to be dependent on both weather conditions and local site effects. Significant seasonal variations in infrasound detections including daily time of occurrence, detection numbers, and phase velocity/azimuth estimates are documented. These time-dependent effects are strongly correlated with atmospheric winds and temperatures and are predicted by available atmospheric specifications. As a result, this suggests that commonly available atmospheric specifications can be used to predict both station and network detection performance, and an appropriate forward model improves location capabilities as a function of time.
47 OTHER INSTRUMENTATION↗
Positional Tolerancing / Feature(s) of Size Surface Interpretation vs Axis Interpretation.
Abstract not provided.
Bryce, Edwin A.↗
Detection of recurrent pancreatic cancer: value of second-opinion interpretations of cross-sectional images by subspecialized radiologists
Purpose: To investigate the value of second-opinion interpretation of cross-sectional images by subspecialized radiologists to diagnose recurrent pancreatic cancer after surgery. Methods: The IRB approved and issued a waiver of informed consent for this retrospective study. Initial and second-opinion interpretations of 69 consecutive submitted MRI or CT follow-up after pancreatic cancer resection between January 1, 2009 and December 31, 2013 were evaluated by one oncologic imaging radiologist, who was blinded to patient’s clinical details and histopathologic data. The reviewer was asked to classify each interpretation in reference of the diagnosis of PDAC recurrence. It was also recorded if the radiologic interpretation recommended additional imaging studies to confirm recurrence. The diagnosis of recurrence was determined by pathology when available, otherwise by imaging follow-up, clinical, or laboratory assessments. Cohen’s kappa statistic was used to assess agreement between initial and second-opinion interpretations. The differences between the initial and second-opinion interpretations were examined using McNemar test or Bowker’s test of symmetry. Results: Disagreement on recurrence between the initial report and the second-opinion interpretation was observed in 32% of cases (22/69; k = 0.44). Second-opinion interpretations had a higher sensitivity and a higher specificity on recurrence compared to the initial interpretations (0.93 vs. 0.75 and 0.90 vs. 0.68, respectively), and the difference in specificity was significant (p = 0.016). Additional imaging studies were recommended more frequently in the initial interpretation (22% vs. 6%, p = 0.006). Conclusions: Our study shows the second-opinion interpretation by subspecialized radiologists improves the detection of pancreatic cancer recurrence after surgical resection.
62 RADIOLOGY AND NUCLEAR MEDICINE↗
Definitions, methods, and applications in interpretable machine learning
Machine-learning models have demonstrated great success in learning complex patterns that enable them to make predictions about unobserved data. In addition to using models for prediction, the ability to interpret what a model has learned is receiving an increasing amount of attention. However, this increased focus has led to considerable confusion about the notion of interpretability. In particular, it is unclear how the wide array of proposed interpretation methods are related and what common concepts can be used to evaluate them. We aim to address these concerns by defining interpretability in the context of machine learning and introducing the predictive, descriptive, relevant (PDR) framework for discussing interpretations. The PDR framework provides 3 overarching desiderata for evaluation: predictive accuracy, descriptive accuracy, and relevancy, with relevancy judged relative to a human audience. Moreover, to help manage the deluge of interpretation methods, we introduce a categorization of existing techniques into model-based and post hoc categories, with subgroups including sparsity, modularity, and simulatability. To demonstrate how practitioners can use the PDR framework to evaluate and understand interpretations, we provide numerous real-world examples. These examples highlight the often underappreciated role played by human audiences in discussions of interpretability. Finally, based on our framework, we discuss limitations of existing methods and directions for future work. We hope that this work will provide a common vocabulary that will make it easier for both practitioners and researchers to discuss and choose from the full range of interpretation methods.
97 MATHEMATICS AND COMPUTING↗
In comparison with other abdominal imaging modalities, which radiologists interpret abdominal MRI?
Purpose: To assess subspecialty mix and case volumes of general and abdominal subspecialty radiologists interpreting abdominal MRI. Methods: The 2016 CMS Physician/Supplier Procedure Summary Master File was used to obtain billed counts of radiologist-interpreted abdominal fluoroscopy, US, CT, and MRI examinations. The CMS Physician and Other Supplier Public Use File was used to assess the subspecialty mix and case volume of the radiologists interpreting those examinations. Results: The fraction of all abdominal imaging examinations interpreted by generalists and abdominal subspecialty radiologists was 70.7% and 16.5% for fluoroscopy; 68.7% and 21.0% for US; 71.4% and 19.2% for CT; and 41.9% and 52.5% for MRI. In 2016, the fraction of general and abdominal radiologists interpreting > 50 fluoroscopy examinations on Medicare fee-for-service beneficiaries was 15.1% and 16.2%. For > 50 US examinations, the fraction was 61.5% and 60.5%; for > 50 CT examinations, 91.2% and 79.6%; and for > 50 MRI examinations, 4.0% and 28.5%. The fraction of abdominal imaging examinations interpreted overall by low-volume providers (those interpreting ≤ 50 examinations in 2016) was 59.5% for fluoroscopy, 17.5% for US, 6.3% for CT, and 50.6% for MRI. Conclusion: Nationally, most abdominal fluoroscopy, US, and CT examinations are interpreted by general radiologists, who have similar annual volumes of these examinations as abdominal subspecialty radiologists. In contrast, most abdominal MRI examinations are interpreted by abdominal subspecialty radiologists, who attain considerably higher volumes. These findings have implications for workforce planning and abdominal imaging fellowship design to ensure their graduates are optimally prepared to contribute to their future practices.
62 RADIOLOGY AND NUCLEAR MEDICINE↗
Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability
Abstract Neural networks have become increasingly prevalent within the geosciences, although a common limitation of their usage has been a lack of methods to interpret what the networks learn and how they make decisions. As such, neural networks have often been used within the geosciences to most accurately identify a desired output given a set of inputs, with the interpretation of what the network learns used as a secondary metric to ensure the network is making the right decision for the right reason. Neural network interpretation techniques have become more advanced in recent years, however, and we therefore propose that the ultimate objective of using a neural network can also be the interpretation of what the network has learned rather than the output itself. We show that the interpretation of neural networks can enable the discovery of scientifically meaningful connections within geoscientific data. In particular, we use two methods for neural network interpretation called backward optimization and layerwise relevance propagation, both of which project the decision pathways of a network back onto the original input dimensions. To the best of our knowledge, LRP has not yet been applied to geoscientific research, and we believe it has great potential in this area. We show how these interpretation techniques can be used to reliably infer scientifically meaningful information from neural networks by applying them to common climate patterns. These results suggest that combining interpretable neural networks with novel scientific hypotheses will open the door to many new avenues in neural network‐related geoscience research.
58 GEOSCIENCES↗
Effectiveness of an e-Learning Platform for Image Interpretation Education of Medical Staff and Students
Medical staff must be able to perform accurate initial interpretations of radiography to prevent diagnostic errors. Education in medical image interpretation is an ongoing need that is addressed by text-based and e-learning platforms. The effectiveness of these methods has been previously reported. Here, we describe the effectiveness of an e-learning platform used for medical image interpretation education. Ten third-year medical students without previous experience in chest radiography interpretation were provided with e-learning instructions. Accuracy of diagnosis using chest radiography was provided before and after e-learning education. We measured detection accuracy for two image groups: nodular shadow and ground-glass shadow. We also distributed the e-learning system to the two groups and analyzed the effectiveness of education for both types of image shadow. The mean correct answer rate after the 2-week e-learning period increased from 34.5 to 72.7%. Diagnosis of the ground glass shadow improved significantly more than that of the mass shadow. Education using the e-leaning platform is effective for interpretation of chest radiography results. E-learning is particularly effective for the interpretation of chest radiography images containing ground glass shadow.
62 RADIOLOGY AND NUCLEAR MEDICINE↗
Identifying geological structures through microseismic cluster and burst analyses complementing active seismic interpretation
At the Decatur carbon capture and storage site (IL, USA) CO 2 has been injected from 2011–2014 and from 2017 to present near the base of the Lower Mt. Simon Sandstone saline reservoir, resulting in microseismicity. Microseismicity is mainly located in the basement and distributed in distinct spatial clusters. The lack of significant impedance contrasts within the basement makes the interpretation of active-source seismic reflection data challenging, however, recent reprocessing allowed to resolve faults above and at the top of the basement. These faults generally do not coincide with the location of microseismic events and their continuation to the general depth of the seismic events cannot be assumed. This paper shows how the interpretation of the microseismicity can complement structural interpretations of active-source seismic reflection data. In particular, we analyze clusters and bursts (abrupt increases) of microseismicity, identify unresolved, smaller-scale weaknesses and extract statistical parameters. These parameters allow comparisons with the interpreted faults, and with fracture sets intercepted by boreholes. During injection at the Decatur site, the injection pressure was kept far below fracture pressure, nevertheless, seismic events were induced and spread far beyond the expected extent of the CO 2 plume. We argue that local stress transfers related to the CO 2 injection reactivated pre-existing fractures within the critically stressed basement. Finally, we conducted a slip tendency analysis for faults interpreted from active seismic, selected cluster, bursts and nodal planes from focal mechanisms to determine if the interpreted structures are optimally oriented with respect to the stress regime. Our results suggest that the orientation of fractures close to the injection well, generally shows slight deviations from the optimal orientation for slip. This might indicate either slight local deviations of the maximum horizontal stress azimuth from the average direction used in the analysis, or the lack of optimally oriented fractures at this location.
58 GEOSCIENCES↗
A Bayesian interpretation of inconsistency measures in cosmology
Abstract Measures of inconsistency and tension between datasets have become an essential part of cosmological analyses. It is important to accurately evaluate the significance of such tensions when present. We propose here a Bayesian interpretation of inconsistency measures that can extract information about physical inconsistencies in the presence of data scatter. This new framework is based on the conditional probability distribution of the level of physical inconsistency given the obtained value of the measure. We use the index of inconsistency as a case study to illustrate the new interpretation framework, but this can be generalized to other metrics. Importantly, there are two aspects in the quantification of inconsistency that behave differently as the number of model parameters increases. The first is the probability for the level of physical inconsistency to reach a threshold which drops with the increase of the number of parameters under consideration. The second is the actual level of physical inconsistency which remains rather insensitive to such an increase in parameters. The difference between these two aspects is often overlooked, which leads to a long-standing ambiguity: when a given inconsistency is found between two constraints, its “significance” seems to be lower when considered in a higher-dimensional parameter space. This ambiguity is resolved by the Bayesian interpretation we introduce in this work because the conditional probability distribution includes all the statistical information of the level of physical inconsistency. Finally, we apply the Bayesian interpretation to examine the (in)consistency between Planck versus the Cepheid-based local measurement, the Dark Energy Survey (DES), the Atacama Cosmology Telescope (ACT) and WMAP. We confirm and revisit the degrees of previous physical inconsistencies and show the stability of the new interpretation with respect to the number of cosmological parameters compared to the commonly used n-σ interpretation when applied to cosmological tensions in multi-parameter spaces.
Astronomy & Astrophysics↗
Balancing Trade-offs: Adaptive Differential Privacy in Interpretable Machine Learning Models
In the advancing field of machine learning, balancing accuracy, interpretability, and privacy represents a significant challenge. The problem is exacerbated by the widespread deployment of pre-trained models locally in diverse applications, which could lead to various amounts of privacy leakage. Conventional Differential Privacy strategies, in which uniform noises are applied to model gradients, guarantee data privacy at the expense of accuracy and interpretability. This paper introduces a Feature-Sensitive Adaptive Differential Privacy (FADP) framework with a unique noise-adding strategy. Noises are adaptively added based on feature importance clustering, where important features are considered for interpretability. By employing a unique masking technique, FADP selectively preserves crucial features with minimal noise interference, maintaining accuracy while enhancing interpretability. The FADP framework addresses the limitations of traditional DP methods by preserving critical channels and improving interpretability — a vital requirement in machine learning applications that demand transparency in model decisions. Through comprehensive testing, FADP is shown to balance the trade-offs among accuracy, privacy, and interpretability, marking a substantial advancement in the field of privacy-preserving machine learning.
Farhad Riya, Farhin [University of Tennessee, Knox↗
Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics
Data-driven constitutive modeling with neural networks has received increased interest in recent years due to its ability to easily incorporate physical and mechanistic constraints and to overcome the challenging and time-consuming task of formulating phenomenological constitutive laws that can accurately capture the observed material response. However, even though neural network-based constitutive laws have been shown to generalize proficiently, the generated representations are not easily interpretable due to their high number of trainable parameters. Sparse regression approaches exist that allow for obtaining interpretable expressions, but the user is tasked with creating a library of model forms which by construction limits their expressiveness to the functional forms provided in the libraries. Here, in this work, we propose to train regularized physics-augmented neural network-based constitutive models utilizing a smoothed version of $L^0$-regularization. This aims to maintain the trustworthiness inherited by the physical constraints, but also enables interpretability which has not been possible thus far on any type of machine learning-based constitutive model where model forms were not assumed a priori but were actually discovered. During the training process, the network simultaneously fits the training data and penalizes the number of active parameters, while also ensuring constitutive constraints such as thermodynamic consistency. We show that the method can reliably obtain interpretable and trustworthy constitutive models for compressible and incompressible hyperelasticity, yield functions, and hardening models for elastoplasticity, using synthetic and experimental data. This work aims to set a new paradigm for interpretable machine learning models in the broad area of solid mechanics where low and limited data is available along with prior knowledge of physical constraints that the learned maps need to obey. This paradigm can potentially be extended to a broader spectrum of scientific exploration.
Data-driven constitutive models↗
Precise interpretations of traditional fine-tuning measures
We uncover two precise interpretations of traditional electroweak fine-tuning (FT) measures that were historically missed. (i) a statistical interpretation : the traditional FT measure shows the change in plausibility of a model in which a parameter was exchanged for the 𝑍 boson mass relative to an untuned model in light of the 𝑍 boson mass measurement. (ii) an information-theoretic interpretation : the traditional FT measure shows the exponential of the extra information, measured in nats, relative to an untuned model that you must supply about a parameter in order to fit the 𝑍 mass. We derive the mathematical results underlying these interpretations, and explain them using examples from weak scale supersymmetry. These new interpretations allow us to rigorously define FT in particle physics and beyond, shed fresh light on the status of extensions to the Standard Model and, lastly, allow us to precisely reinterpret historical and recent studies using traditional FT measures.
electroweak symmetry breaking↗