SEARCH · Search NASA

Results for “DATA SMOOTHING”

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 37 records · Page 2

All sources NASA NTRS DOE OSTI

DOE OSTI2020

Quantifying the generalization error in deep learning in terms of data distribution and neural network smoothness

We report the accuracy of deep learning, i.e., deep neural networks, can be characterized by dividing the total error into three main types: approximation error, optimization error, and generalization error. Whereas there are some satisfactory answers to the problems of approximation and optimization, much less is known about the theory of generalization. Most existing theoretical works for generalization fail to explain the performance of neural networks in practice. To derive a meaningful bound, we study the generalization error of neural networks for classification problems in terms of data distribution and neural network smoothness. We introduce the cover complexity (CC) to measure the difficulty of learning a data set and the inverse of the modulus of continuity to quantify neural network smoothness. A quantitative bound for expected accuracy/error is derived by considering both the CC and neural network smoothness. Although most of the analysis is general and not specific to neural networks, we validate our theoretical assumptions and results numerically for neural networks by several data sets of images. The numerical results confirm that the expected error of trained networks scaled with the square root of the number of classes has a linear relationship with respect to the CC. We also observe a clear consistency between test loss and neural network smoothness during the training process. In addition, we demonstrate empirically that the neural network smoothness decreases when the network size increases whereas the smoothness is insensitive to training dataset size.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Calculating the grain boundary inclination of voxelated grain structures using a smoothing algorithm

We have developed a flexible method for calculating the grain boundary (GB) inclinations of voxelated grain structure data using smoothing algorithms. We compared the performance of four algorithms: the linear interpolation, Allen–Cahn, level-set, and vertex algorithms. We assessed their accuracy using 2D and 3D cases with known inclinations. The vertex algorithm provided the best balance between accuracy and efficiency for 2D structures while the linear interpolation algorithm provided the best balance for 3D structures. We compared the GB inclinations calculated using our smoothing method on a 3D high energy X-ray diffraction microscopy (HEDM) dataset to those determined by meshing the GBs. The two approaches determined similar GB plane distributions, though they varied significantly at triple junctions. In conclusion, the smoothing method was demonstrated for two sources of 3D voxelated grain structures: HEDM data and results from Monte Carlo Potts grain growth simulations.

36 MATERIALS SCIENCE↗

DOE OSTI2022

Metrics for Intercomparison of Remapping Algorithms (MIRA) protocol applied to Earth system models

Abstract. Strongly coupled nonlinear phenomena such as those described by Earth system models (ESMs) are composed of multiple component models with independent mesh topologies and scalable numerical solvers. A common operation in ESMs is to remap or interpolate component solution fields defined on their computational mesh to another mesh with a different combinatorial structure and decomposition, e.g., from the atmosphere to the ocean, during the temporal integration of the coupled system. Several remapping schemes are currently in use or available for ESMs. However, a unified approach to compare the properties of these different schemes has not been attempted previously. We present a rigorous methodology for the evaluation and intercomparison of remapping methods through an independently implemented suite of metrics that measure the ability of a method to adhere to constraints such as grid independence, monotonicity, global conservation, and local extrema or feature preservation. A comprehensive set of numerical evaluations is conducted based on a progression of scalar fields from idealized and smooth to more general climate data with strong discontinuities and strict bounds. We examine four remapping algorithms with distinct design approaches, namely ESMF Regrid (Hill et al., 2004), TempestRemap (Ullrich and Taylor, 2015), generalized moving least squares (GMLS) (Trask and Kuberry, 2020) with post-processing filters, and WLS-ENOR (Li et al., 2020). By repeated iterative application of the high-order remapping methods to the test fields, we verify the accuracy of each scheme in terms of their observed convergence order for smooth data and determine the bounded error propagation using challenging, realistic field data on both uniform and regionally refined mesh cases. In addition to retaining high-order accuracy under idealized conditions, the methods also demonstrate robust remapping performance when dealing with non-smooth data. There is a failure to maintain monotonicity in the traditional L2-minimization approaches used in ESMF and TempestRemap, in contrast to stable recovery through nonlinear filters used in both meshless GMLS and hybrid mesh-based WLS-ENOR schemes. Local feature preservation analysis indicates that high-order methods perform better than low-order dissipative schemes for all test cases. The behavior of these remappers remains consistent when applied on regionally refined meshes, indicating mesh-invariant implementations. The MIRA intercomparison protocol proposed in this paper and the detailed comparison of the four algorithms demonstrate that the new schemes, namely GMLS and WLS-ENOR, are competitive compared to standard conservative minimization methods requiring computation of mesh intersections. The work presented in this paper provides a foundation that can be extended to include complex field definitions, realistic mesh topologies, and spectral element discretizations, thereby allowing for a more complete analysis of production-ready remapping packages.

58 GEOSCIENCES↗

DOE OSTI2020

Vector Field Decompositions Using Multiscale Poisson Kernel

Extraction of multiscale features using scale-space is one of the fundamental approaches to analyze scalar fields. However, similar techniques for vector fields are much less common, even though it is well known that, for example, turbulent flows contain cascades of nested vortices at different scales. The challenge is that the ideas related to scale-space are based upon iteratively smoothing the data to extract features at progressively larger scale, making it difficult to extract overlapping features. Instead, we consider spatial regions of influence in vector fields as scale, and introduce a new approach for the multiscale analysis of vector fields. Rather than smoothing the flow, we use the natural Helmholtz-Hodge decomposition to split it into small-scale and large-scale components using progressively larger neighborhoods. Our approach creates a natural separation of features by extracting local flow behavior, for example, a small vortex, from large-scale effects, for example, a background flow. We demonstrate our technique on large-scale, turbulent flows, and show multiscale features that cannot be extracted using state-of-the-art techniques.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2026

Quantum fate of the Choptuik naked singularity

Classical critical collapse provides a dynamical route from smooth initial data to a naked singularity, representing a sharper violation of predictability than ordinary black hole singularities. We argue that this distinction is erased by quantum backreaction. Building on the semiclassical interior analysis, where quantum self-energy of the collapsing matter generates a universal growing mode and a finite mass gap, we study the exterior naked singularity region that determines global visibility in the Einstein-scalar system. We analyze controlled exterior models in both $2+1$ and $3+1$ dimensions. In the former, smooth matching and physical boundary conditions analytically select a vacuum polarization state, whose backreaction cloaks the classically naked region by a quantum trapped branch. In the latter, numerical horizon tracing shows that near a quantum-shifted threshold the exterior develops finite-mass marginally trapped surfaces rather than a zero-mass naked endpoint. These results suggest a global quantum picture in which the Choptuik naked singularity shares the fate of an ordinary black hole singularity: quantum effects push the putative Cauchy horizon behind a quantum-generated horizon, thereby reducing the loss of predictability to the standard black hole evaporation problem.

FOS: Physical sciences↗

DOE OSTI2022

MatLab Package for Whispering Gallery Mode Data

Dye-doped whispering gallery mode resonator (WGMR) microspheres yield highly structured emission spectra that are extremely sensitive to their environment and are of intense interest for use in a variety of sensing applications. Efforts to leverage the unique sensitivities of WGMRs have relied on stringent experimental requirements to correlate specific spectral shifts/changes to an analyte/stimulus such as 1) precise positional knowledge, 2) reference spectra for each microsphere, and 3) high mechanical stability. Consequently, these can hinder adequate mixing or incorporation of analytes and creates challenges for remote sensing. The MATLAB codes provided here are to be used in conjunction with a continuous flow technique for measuring WGM spectra of dye-doped microspheres suspended in solution. One MATLAB script, smooths the data, automatically baseline corrects it to isolate the whispering gallery modes (WGM) from the unwanted bulk emission, and assesses the similarity of each spectrum to aid in selecting a set of unique WGM spectra for further analysis. The next script is designed to analyze WGM spectra to determine the size of the resonator and the refractive index (RI) of its local environment without a priori knowledge of the individual microsphere. The final script allows the user to cluster spheres based on the product of their RI and radius and the contrast ratio of the RI of the sphere material and that of its environment by using a shared nearest neighbor spectral clustering algorithm.

Lilley, Laura↗

DOE OSTI2026

Unveiling horizons in quantum critical collapse

Critical gravitational collapse offers a unique window into regimes of arbitrarily high curvature, culminating in a naked singularity arising from smooth initial data — thus providing a dynamical counterexample to weak cosmic censorship. Near the critical regime, quantum effects from the collapsing matter are expected to intervene before full quantum gravity resolves the singularity. Despite its fundamental significance, a self-consistent treatment has so far remained elusive. In this work, we perform a one-loop semiclassical analysis using the robust anomaly-based method in the canonical setup of Einstein gravity minimally coupled to a free, massless scalar field. Focusing on explicitly solvable near-critical solutions in both 2 + 1 and 3 + 1 dimensions, we analytically solve the semiclassical Einstein equations and obtain controlled, quantitative results for several long-standing questions within the dominant s-wave sector. We find that regularity uniquely selects a Boulware-like quantum state, encoding genuine vacuum polarization effects from the collapsing matter. Remarkably, the resulting quantum corrections manifest as a growing mode. Horizon-tracing analyses, incorporating both classical and quantum modes, reveal the emergence of a finite mass gap, signaling a phase transition from classical Type II to quantum-modified Type I behavior, thereby providing a quantum enforcement of the weak cosmic censorship. The most non-trivial aspect of our analysis involves dealing with non-conformal matter fields in explicitly time-dependent critical spacetimes. Along the way, we uncover intriguing and previously underexplored features of quantum field theory in curved spacetime.

2D Gravity↗

DOE OSTI2021

Improving the astrometric solution of the Hyper Suprime-Cam with anisotropic Gaussian processes

Context. We study astrometric residuals from a simultaneous fit of Hyper Suprime-Cam images. Aims. We aim to characterize these residuals and study the extent to which they are dominated by atmospheric contributions for bright sources. Methods. We used Gaussian process interpolation with a correlation function (kernel) measured from the data to smooth and correct the observed astrometric residual field. Results. We find that a Gaussian process interpolation with a von Kármán kernel allows us to reduce the covariances of astrometric residuals for nearby sources by about one order of magnitude, from 30 mas 2 to 3 mas 2 at angular scales of ~1 arcmin. This also allows us to halve the rms residuals. Those reductions using Gaussian process interpolation are similar to recent result published with the Dark Energy Survey dataset. We are then able to detect the small static astrometric residuals due to the Hyper Suprime-Cam sensors effects. We discuss how the Gaussian process interpolation of astrometric residuals impacts galaxy shape measurements, particularly in the context of cosmic shear analyses at the Rubin Observatory Legacy Survey of Space and Time.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2026

Quantum critical collapse abhors a naked singularity

Classical critical collapse yields naked singularities from smooth initial data, challenging cosmic censorship, and shaping the spectrum of primordial black holes. We show that one-loop vacuum polarization near the threshold qualitatively changes this outcome by dressing the singularity with a horizon within a controlled semiclassical regime. In analytically tractable Einstein-scalar critical spacetimes, a one-loop $s$-wave treatment linearized around self-similar backgrounds shows that regularity uniquely selects an asymptotically Minkowskian, vacuum-polarization state. Its renormalized stress tensor carries a universal quantum growing mode that competes with the classical unstable mode, shifts the critical point, and generates a trapped surface along with a finite mass gap at the new threshold, thereby enforcing horizon formation even under arbitrary fine-tuning. In primordial collapse, the threshold shift enters exponentially into the formation fraction, while the mass gap truncates the low-mass tail, suggesting potentially important consequences for the predicted mass spectrum. Furthermore, these results provide a self-consistent semiclassical treatment of critical collapse and yield sharp predictions within the one-loop, near-critical, linearized regime.

Anomalies↗

DOE OSTI2022

Additive Manufacturing Flaw Assessment Methodology

An evolution fatigue data and flaw tolerance of components produced using the Powder Bed Fusion (PBF) Additive Manufacturing (AM) process is documented in this report. Initial differences in fatigue data for AM components compared to smooth bar fatigue data indicated a need for a more detailed analysis of AM data available in technical literature. The investigation was initiated to support the development of a fatigue analysis methodology for AM components to support the of codification of AM technology for pressure equipment. The project was initiated to collect and analyze stainless steel 304L and 316L AM fatigue test data and corresponding process and quality information to develop S-N and E-N based fatigue data representation. Additional AM fatigue test data including Inconel Ti-6-4 and aluminum alloys were also considered for comparison purposes Metallic AM parts tend to contain various forms of defects distributed throughout the part. If an AM part is subjected to fatigue loading in service, a fatigue analysis needs to be performed during the design process to ensure an acceptable service life for the part. Post-process machining and polishing do not to improve fatigue resistance in any significant degree. The low cycle fatigue regime is of particular interest to this project in support of flaw acceptance criteria currently under development by ASME’s BPTCS/BNCS. Internal defects become exposed as external surface defects during machining for the machining of the AM part to final dimensions. This implies that as long as inherent AM defects are within a controlled limit in terms of both size and distribution characteristics, the corresponding fatigue test data in terms of either S-N (stress life) or E-N (strain life) can be investigated and characterized to establish fatigue properties of AM parts for design and fatigue evaluation purposes. The resulting S-N or E-N curves and their scatter bands can be used to derive fatigue design allowable stress values by capturing the effects of distributed discontinuities within an acceptable limit.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Leaf phenology data at The Morton Arboretum Forestry Plots 2019-2023

We are collecting long-term leaf phenology data at The Morton Arboretum to determine seasonal patterns of leaf production in trees. This data on leaf phenology will be integrated with other ongoing data streams to create a connection between above- and below-ground tree processes. This data package contains raw and smooth outputs from phenology data, as well as extracted phenophase dates (i.e., start, peak, and end of season): the raw and smooth outputs from the PhenoCam GUI can be found in the "leafRaw.csv" and "leafSmooth.csv" files, respectively, and the extracted phenophase dates can be found in the "leafPhenophaseDates2019-2023.csv" file. Extracted phenophase dates for evergreen species in 2023 are currently unavailable, and the files will be updated once they are extracted. Additional information on units and other file-level metadata can be found within each data file's respective data dictionary, and metadata for each of the 23 surveyed plots can be found within the "Location_metadata.csv" file. While the "leafRaw.csv" and the "leafSmooth.csv" files contain all data for all species, the "leafPhenophaseDates2019-2023.csv" file currently excludes the dates for evergreen species in 2023. Another version of the file will be added as those dates are extracted.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Condensation heat transfer and pressure drop of low-global warming potential refrigerants in smooth aluminum tubes

In this work we report condensation heat transfer coefficient and pressure drop data for R-454B, R-32, and R-454C. These refrigerants are low Global Warming Potential (GWP) replacements for R-410A. We obtained these data in a smooth 7.2 mm outer diameter (5.94 mm inner diameter) aluminum tube. In these experiments, the key variables were condensation temperature (35 °C ≤ T cond ≤ 50 °C) and refrigerant mass flux (151.1 ≤ G ≤ 347.3 kg m –2 s –1 ). In general, for similar nominal operating conditions, we measured the highest heat transfer coefficients for R-32 and the lowest for R-454C. Among the correlations we used to predict the heat transfer data for all three refrigerants, the one developed by Thome et al. (2003) predicted the data with the greatest accuracy. Deviations between experimental data and predictions of this correlation were 21.7% for R-454B (with SBG correction), 22.0% for R-32, and 18.8% for R-454C (with SBG correction). Experimental pressure gradient data were predicted by the Friedel correlation Friedel (1979) with deviations of 13.8% for R-454B, 9.8% for R-32, and 10.2% for R-454C. In this work, we also compare the predictions of the Cavallini et al. correlation Cavallini et al. (2006) for R-454B and R-32 in smooth copper versus smooth aluminum tubes. To improve this correlation’s predictive capability for aluminum tubes, we have proposed some modifications. After incorporating these modifications, the deviations between the correlation’s predictions and our data were 10% for R-454B (with SBG correction), 8.1% for R-32, and 5.2% for R-454C (with SBG correction). Furthermore, these data and findings will assist the HVAC and refrigeration industry to implement low-GWP refrigerants in condensers using smooth aluminum tubes.

Aluminum tube↗

DOE OSTI2026

Extracting the Breakout Distance from the ECOT Trajectories: Gaussian Process Regression Approach

Enhanced Corner Turning (ECOT) experiments provide an important metric of performance of high explosive (HE) formulations. The breakout distance is a single scalar value that characterizes the corner turning efficiency of an HE. Extracting the breakout distance from the raw ECOT results, whether experimental or simulated, is a conceptually straightforward procedure which, however, is non-unique, especially in the presence of noise. More specifically, this procedure involves numerical smoothing and selecting particular values for parameters of this smoothing introduces human bias. In this work, we propose to use the Gaussian process regression to analyze ECOT results. This analysis involves the effective smoothing of the data, thus allowing for accurate extraction of the breakout distance. Most importantly, the parameters of this smoothing can be inferred from the ECOT data itself, rendering the approach effectively parameter-free and thus diminishing the human bias. An additional benefit of the Gaussian process regression, being a statistical inference method, is that not just the value of the breakout distance, but also its confidence interval can be extracted from the data. This report introduces the Gaussian process regression, as applied to ECOT, and demonstrates its usefulness by extracting the breakout distances for a selection of experimental and simulated data.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

DOE OSTI2022

Diffuse X-ray scattering from polished silicon: application of the distorted wave Born approximation

Measured diffuse X-ray scattering data for a `smooth' as well as for a `rough' silicon sample were fit to theoretical expressions within the distorted wave Born approximation (DWBA). Data for the power spectral density (PSD) for both samples were also obtained by means of atomic force microscopy and optical interferometry. The Fourier transforms of trial correlation functions were fit to the PSD data and then applied to the DWBA formalism. The net correlation functions needed to fit the PSD data for each sample comprised the sum of two terms with different cutoff lengths and different self-affine fractal exponents. At zero distance these correlation functions added up to yield net values of σ 2 = (2) 2 and (71) 2 Å 2 for the smooth and rough samples, respectively. X-ray scattering data were obtained at beamline 1-BM of the Advanced Photon Source. Data and fits at values of q z = 0.05 and 0.10 Å −1 for the smooth sample are reported. Good fits for the smooth sample were obtained at both q z values simultaneously, that is, identical fitting parameters were applied at both values of q z . The smooth sample also exhibited weak Yoneda wings and a clear distinction between the strong specular scattering and the weak diffuse scattering. Data for the rough sample were qualitatively different and exhibited very weak scattering at the specular condition in contrast to extremely large Yoneda wings. Fits for the rough sample are reported for q z = 0.04, 0.05, and 0.06 Å −1 . Although the large Yoneda wings could be fit quite well in both position and amplitude, scattering near the specular condition could not be equally well fit by applying the same fitting parameters at all values of q z . Albeit imperfect, best-fitting results at the specular condition were obtained by invoking only diffuse scattering, that is, without including a separate theoretical expression for specular scattering.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2020

Study on the High-Pressure Behavior of Goethite up to 32 GPa Using X-Ray Diffraction, Raman, and Electrical Impedance Spectroscopy

Goethite is a major iron-bearing sedimentary mineral on Earth. In this study, we conducted in situ high-pressure x-ray diffraction, Raman, and electrical impedance spectroscopy measurements of goethite using a diamond anvil cell (DAC) at room temperature and high pressures up to 32 GPa. We observed feature changes in both the Raman spectra and electrical resistance at about 5 and 11 GPa. However, the x-ray diffraction patterns show no structural phase transition in the entire pressure range of the study. The derived pressure-volume (P-V) data show a smooth compression curve with no clear evidence of any second-order phase transition. Fitting the volumetric data to the second-order Birch–Murnaghan equation of state yields V 0 = 138.9 ± 0.5 Å 3 and K 0 = 126 ± 5 GPa.

58 GEOSCIENCES↗

DOE OSTI2023

A General Spatiotemporal Imputation Framework for Missing Sensor Data

Many applications from precision agriculture, environmental monitoring and transportation networks rely on data collected across space and time over a large geographic area. Missing data poses a significant challenge for any data-driven inference and control tasks. Data imputation or the estimation of missing data can help fill these gaps by utilizing inherent spatial relationships and temporal patterns. A variety of spatiotemporal imputation models have been developed to address missing data in spatiotemporal datasets. However, these classical methods rely on the assumption that the underlying data follows a smooth trend and fail to provide accurate estimates when there is a large number of missing points in the data. Even though there are machine learning driven tensor completion approaches such as convolutional neural network based tensor completion (CoSTCo) that capture the non-linear relationships in the dataset, the transductive nature makes the algorithm less scalable. Thus, existing approaches for estimating the missing information do not effectively capture all dimensions of the spatiotemporal data structure, resulting in erroneous predictions and poor performance. The main contributions of this paper are: (1) We propose a novel inductive framework (G-LSTM) for missing data imputation that integrates a graph neural network with LSTMs to effectively capture both spatial and temporal dependencies. (2) Experimental results on a traffic dataset demonstrate that the proposed GNN integrated with an LSTM framework achieves improved imputation and maintains steady performance even when there are extreme missing conditions in comparison with the state-of-the-art imputation framework (i.e, CoSTCo). (3) The simulation results on a traffic network show up to 69% reduction in mean absolute error and 61% reduction in root mean square error when compared to CoSTCo.

Tharzeen, Aabila↗

DOE OSTI2020

Planck intermediate results

We present the NPIPE processing pipeline, which produces calibrated frequency maps in temperature and polarization from data from the Planck Low Frequency Instrument (LFI) and High Frequency Instrument (HFI) using high-performance computers. NPIPE represents a natural evolution of previous Planck analysis efforts, and combines some of the most powerful features of the separate LFI and HFI analysis pipelines. For example, following the LFI 2018 processing procedure, NPIPE uses foreground polarization priors during the calibration stage in order to break scanning-induced degeneracies. Similarly, NPIPE employs the HFI 2018 time-domain processing methodology to correct for bandpass mismatch at all frequencies. In addition, NPIPE introduces several improvements, including, but not limited to: inclusion of the 8% of data collected during repointing manoeuvres; smoothing of the LFI reference load data streams; in-flight estimation of detector polarization parameters; and construction of maximally independent detector-set split maps. For component-separation purposes, important improvements include: maps that retain the CMB Solar dipole, allowing for high-precision relative calibration in higher-level analyses; well-defined single-detector maps, allowing for robust CO extraction; and HFI temperature maps between 217 and 857 GHz that are binned into 0'.9 pixels (Nside = 4096), ensuring that the full angular information in the data is represented in the maps even at the highest Planck resolutions. The net effect of these improvements is lower levels of noise and systematics in both frequency and component maps at essentially all angular scales, as well as notably improved internal consistency between the various frequency channels. Based on the NPIPE maps, we present the first estimate of the Solar dipole determined through component separation across all nine Planck frequencies. The amplitude is (3366.6 ± 2.7) μK, consistent with, albeit slightly higher than, earlier estimates. From the large-scale polarization data, we derive an updated estimate of the optical depth of reionization of τ = 0.051 ± 0.006, which appears robust with respect to data and sky cuts. There are 600 complete signal, noise and systematics simulations of the full-frequency and detector-set maps. As a Planck first, these simulations include full time-domain processing of the beam-convolved CMB anisotropies. The release of NPIPE maps and simulations is accompanied with a complete suite of raw and processed time-ordered data and the software, scripts, auxiliary data, and parameter files needed to improve further on the analysis and to run matching simulations.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2024

Calibrating Bayesian generative machine learning for Bayesiamplification

Recently, combinations of generative and Bayesian deep learning have been introduced in particle physics for both fast detector simulation and inference tasks. These neural networks aim to quantify the uncertainty on the generated distribution originating from limited training statistics. The interpretation of a distribution-wide uncertainty however remains ill-defined. We show a clear scheme for quantifying the calibration of Bayesian generative machine learning models. For a Continuous Normalizing Flow applied to a low-dimensional toy example, we evaluate the calibration of Bayesian uncertainties from either a mean-field Gaussian weight posterior, or Monte Carlo sampling network weights, to gauge their behaviour on unsteady distribution edges. Well calibrated uncertainties can then be used to roughly estimate the number of uncorrelated truth samples that are equivalent to the generated sample and clearly indicate data amplification for smooth features of the distribution.

97 MATHEMATICS AND COMPUTING↗