Search NASA⌕ Search

SEARCH · Search NASA

Results for “constrained estimation”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

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

At least 307 records · Page 17
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Analysis of the Trusted Inertial Terrain-Aided Navigation Measurement Function

The trusted inertial terrain-aided navigation (TITAN) algorithm leverages an airborne vertical synthetic aperture radar to measure the range to the closest ground points along several prescribed iso-Doppler contours. These TITAN minimum-range, prescribed-Doppler measurements are the result of a constrained nonlinear optimization problem whose optimization function and constraints both depend on the radar position and velocity. Owing to the complexity of this measurement definition, analysis of the TITAN algorithm is lacking in prior work. This publication offers such an analysis, making the following three contributions: (1) an analytical solution to the TITAN constrained optimization measurement problem, (2) a derivation of the TITAN measurement function Jacobian, and (3) a derivation of the Cramér-Rao lower bound on the estimated position and velocity error covariance. These three contributions are verified via Monte Carlo simulations over synthetic terrain, which further reveal two remarkable properties of the TITAN algorithm: (1) the along-track positioning errors tend to be smaller than the cross-track positioning errors, and (2) the cross-track positioning errors are independent of the terrain roughness.

TITAN↗
DOE OSTI2022

A Compound Poisson Generator Approach to Point-source Inference in Astrophysics

Abstract The identification and description of point sources is one of the oldest problems in astronomy, yet even today the correct statistical treatment for point sources remains one of the field’s hardest problems. For dim or crowded sources, likelihood-based inference methods are required to estimate the uncertainty on the characteristics of the source population. In this work, a new parametric likelihood is constructed for this problem using compound Poisson generator (CPG) functionals that incorporate instrumental effects from first principles. We demonstrate that the CPG approach exhibits a number of advantages over non-Poissonian template fitting (NPTF)—an existing method—in a series of test scenarios in the context of X-ray astronomy. These demonstrations show that the effect of the point-spread function, effective area, and choice of point-source spatial distribution cannot, generally, be factorized as they are in NPTF, while the new CPG construction is validated in these scenarios. Separately, an examination of the diffuse-flux emission limit is used to show that most simple choices of priors on the standard parameterization of the population model can result in unexpected biases: when a model comprising both a point-source population and diffuse component is applied to this limit, nearly all observed flux will be assigned to either the population or to the diffuse component. A new parameterization is presented for these priors that properly estimates the uncertainties in this limit. In this choice of priors, CPG correctly identifies that the fraction of flux assigned to the population model cannot be constrained by the data.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Primordial non-Gaussianities with weak lensing: information on non-linear scales in the Ulagam full-sky simulations

Abstract Primordial non-Gaussianities (PNGs) are signatures in the density field that encode particle physics processes from the inflationary epoch. Such signatures have been extensively studied using the Cosmic Microwave Background, through constraining their amplitudes,f X NL , with future improvements expected from large-scale structure surveys; specifically, the galaxy correlation functions. We show that weak lensing fields can be used to achieve competitive and complementary constraints. This is shown via theUlagamsuite of N-body simulations, a subset of which evolves primordial fields with four types of PNGs. We create full-sky lensing maps and estimate the Fisher information from three summary statistics measured on the maps: the moments, the cumulative distribution function, and the 3-point correlation function. We find that the year 10 sample from the Rubin Observatory Legacy Survey of Space and Time (LSST) can constrain PNGs toσ(f NL eq ) ≈ 110,σ(f NL or, lss ) ≈ 120,σ(f NL loc ) ≈ 40. For the former two, this is better than or comparable to expected galaxy clustering-based constraints from the Dark Energy Spectroscopic Instrument (DESI). The PNG information in lensing fields is on non-linear scales and at low redshifts (z≲ 1.25), with a clear origin in the evolution history of massive halos. The constraining power degrades by ∼60% under scale cuts of ≳ 20 Mpc, showing there is still significant information on scales mostly insensitive to small-scale systematic effects (e.g., baryons). We publicly release theUlagamsuite to enable more survey-focused analyses.

Astronomy & Astrophysics↗
DOE OSTI2025

Constrained or unconstrained? Neural-network-based equation discovery from data

Throughout many fields, practitioners often rely on differential equations to model systems. Yet, for many applications, the theoretical derivation of such equations and/or the accurate resolution of their solutions may be intractable. Instead, recently developed methods, including those based on parameter estimation, operator subset selection, and neural networks, allow for the data-driven discovery of both ordinary and partial differential equations (PDEs), on a spectrum of interpretability. The success of these strategies is often contingent upon the correct identification of representative equations from noisy observations of state variables and, as importantly and intertwined with that, the mathematical strategies utilized to enforce those equations. Specifically, the latter has been commonly addressed via unconstrained optimization strategies. Representing the PDE as a neural network, we propose to discover the PDE (or the associated operator) by solving a constrained optimization problem and using an intermediate state representation similar to a physics-informed neural network (PINN). The objective function of this constrained optimization problem promotes matching the data, while the constraints require that the discovered PDE is satisfied at a number of spatial collocation points. We present a penalty method and a widely used trust-region barrier method to solve this constrained optimization problem, and we compare these methods on numerical examples. Our results on several example problems demonstrate that the latter constrained method outperforms the penalty method, particularly for higher noise levels or fewer collocation points. This work motivates further exploration into using sophisticated constrained optimization methods in scientific machine learning, as opposed to their commonly used, penalty-method or unconstrained counterparts. For both of these methods, we solve these discovered neural network PDEs with classical methods, such as finite difference methods, as opposed to PINNs-type methods relying on automatic differentiation. Here, we briefly highlight how simultaneously fitting the data while discovering the PDE improves the robustness to noise and other small, yet crucial, implementation details.

Data-driven discovery↗
DOE OSTI2023

Physics-based Induced Earthquake Forecasting: Process Understanding, and Hazards Mitigation

Disposal of saltwater co-produced with oil and gas is linked to elevated seismicity in the Central and Midwest US. There is a concern that these events may lead to widespread damage and an overall increase in seismicity. Thus an improved understanding of the spatially and temporally variable deformation and stress field associated with fluid injection operation is critically necessary for evaluating time-varying seismic hazards. Despite the improvements in seismic monitoring capacity and the resulting decrease in the magnitude detection threshold, estimates of induced earthquake probability remain elusive due to insufficient models incapable of accounting for the complex physics governing the process of induced seismicity. The proposed research effort will comprehensively analyze, integrate, and interpret geodetic, injection, and seismic data in the vicinity of the injection sites in Oklahoma to resolve the 4-dimensional distribution of pore pressure and stress in the shallow crust. This project, in particular, is focused on exploring the statistical relation between injection operation and increased earthquake hazard. The amplitude of and the extent to which pore pressure changes are determined by some factors, in particular, the hydrogeological properties of the rocks, such as diffusivity. Thus the available deformation data will be used to constrain hydrogeological properties of the medium, to accurately resolve the evolution of crustal stresses due to fluid injection. Having the time-varying models of stress changes, a statistical framework will be implemented to estimate the time-dependent probability of large earthquakes on the nearby fault systems. These data and models help to improve seismic hazard estimates and aid in constructing operational-induced earthquake forecast models. This information can also be integrated into the updated U.S. National Seismic Hazard Map, which local communities and authorities use in their earthquake risk estimates and mitigation efforts.

58 GEOSCIENCES↗
DOE OSTI2019

First-Principles Calculation of Water p K a Using the Newly Developed SCAN Functional

Acid/base chemistry is an intriguing topic that still constitutes a challenge for computational chemistry. While estimating the acid dissociation constant (or pK a ) could shed light on many chemistry processes, especially in the fields of biochemistry and geochemistry, evaluating the relative stability between protonated and nonprotonated species is often very difficult. Indeed, a prerequisite for calculating the pK a of any molecule is an accurate description of the energetics of water dissociation. In this study, we applied constrained molecular dynamics simulations, a noncanonical sampling technique, to investigate the water deprotonation process by selecting the OH distance as the reaction coordinate. The calculation is based on density functional theory and the newly developed SCAN functional, which has shown excellent performance in describing water structure. This first benchmark of SCAN on a chemical reaction shows that this functional accurately models the energetics of proton transfer reactions in an aqueous environment. After taking Coulomb long-range corrections and nuclear quantum effects into account, the estimated water pK a is only 1.0 pK a unit different from the target experimental value. Our results show that the combination of SCAN and constrained MD successfully reproduces the chemistry of water and constitutes a good framework for calculating the free energy of chemical reactions of interest.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

A Bayesian Approach for Statistical–Physical Bulk Parameterization of Rain Microphysics. Part II: Idealized Markov Chain Monte Carlo Experiments

Observationally informed development of a new framework for bulk rain microphysics, the Bayesian Observationally Constrained Statistical–Physical Scheme (BOSS; described in Part I of this study), is demonstrated. This scheme’s development is motivated by large uncertainties in cloud and weather simulations associated with approximations and assumptions in existing microphysics schemes. Here, a proof-of-concept study is introduced using a Markov chain Monte Carlo sampling algorithm with BOSS to probabilistically estimate microphysical process rates and parameters directly from a set of synthetically generated rain observations. The framework utilized is an idealized steady-state one-dimensional column rainshaft model with specified column-top rain properties and a fixed thermodynamical profile. Different configurations of BOSS—flexibility being a key feature of this approach—are constrained via synthetic observations generated from a traditional three-moment bulk microphysics scheme. The ability to retrieve correct parameter values when the true parameter values are known is illustrated. For cases when there is no set of true parameter values, the accuracy of configurations of BOSS that have different levels of complexity is compared. It is found that addition of the sixth moment as a prognostic variable improves prediction of the third moment (proportional to bulk rain mass) and rain rate. In contrast, increasing process rate formulation complexity by adding more power terms has little benefit—a result that is explained using further-idealized experiments. BOSS rainshaft simulations are shown to well estimate the true process rates from constraint by bulk rain observations, with the further benefit of rigorously quantified uncertainty of these estimates.

58 GEOSCIENCES↗
DOE OSTI2020

The consequences of gamma-ray burst jet opening angle evolution on the inferred star formation rate

ABSTRACT Gamma-ray burst (GRB) data suggest that the jets from GRBs in the high redshift universe are more narrowly collimated than those at lower redshifts. This implies that we detect relatively fewer long GRB progenitor systems (i.e. massive stars) at high redshifts, because a greater fraction of GRBs have their jets pointed away from us. As a result, estimates of the star formation rate (SFR; from the GRB rate) at high redshifts may be diminished if this effect is not taken into account. In this paper, we estimate the SFR using the observed GRB rate, accounting for an evolving jet opening angle. We find that the SFR in the early universe (z > 3) can be up to an order of magnitude higher than the canonical estimates, depending on the severity of beaming angle evolution and the fraction of stars that make long GRBs. Additionally, we find an excess in the SFR at low redshifts, although this lessens when accounting for evolution of the beaming angle. Finally, under the assumption that GRBs do, in fact, trace canonical forms of the cosmic SFR, we constrain the resulting fraction of stars that must produce GRBs, again accounting for jet beaming-angle evolution. We find this assumption suggests a high fraction of stars in the early universe producing GRBs – a result that may, in fact, support our initial assertion that GRBs do not trace canonical estimates of the SFR.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Probing the Capability of Future Direct-imaging Missions to Spectrally Constrain the Frequency of Earth-like Planets

A critical question in astrobiology is whether exo-Earth candidates (EECs) are Earth-like, in that they originate life that progressively oxygenates their atmospheres similarly to Earth. We propose answering this question statistically by searching for O{sub 2} and O{sub 3} on EECs with missions such as HabEx or LUVOIR. We explore the ability of these missions to constrain the fraction, f {sub E}, of EECs that are Earth-like in the event of a null detection of O{sub 2} or O{sub 3} on all observed EECs. We use the Planetary Spectrum Generator to simulate observations of EECs with O{sub 2} and O{sub 3} levels based on Earth’s history. We consider four instrument designs—LUVOIR-A (15 m), LUVOIR-B (8 m), HabEx with a starshade (4 m, “HabEx/SS”), and HabEx without a starshade (4 m, “HabEx/no-SS”)—as well as three estimates of the occurrence rate of EECs (η {sub earth}): 24%, 5%, and 0.5%. In the case of a null detection, we find that for η {sub earth} = 24%, LUVOIR-A, LUVOIR-B, and HabEx/SS would constrain f {sub E} to ≤0.094, ≤0.18, and ≤0.56, respectively. This also indicates that if f {sub E} is greater than these upper limits, we are likely to detect O{sub 3} on at least one EEC. Conversely, we find that HabEx/no-SS cannot constrain f {sub E}, due to the lack of a coronagraph ultraviolet channel. For η {sub earth} = 5%, only LUVOIR-A and LUVOIR-B would be able to constrain f {sub E}, to ≤0.45 and ≤0.85, respectively. For η {sub earth} = 0.5%, none of the missions would allow us to constrain f {sub E}, due to the low number of detectable EECs. We conclude that the ability to constrain f {sub E} is more robust to uncertainties in η {sub earth} for missions with larger aperture mirrors. However, all missions are susceptible to an inconclusive null detection if η {sub earth} is sufficiently low.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2022

SPARC: Structural properties associated with residue constraints

SPARC facilitates the generation of plausible hypotheses regarding underlying biochemical mechanisms by structurally characterizing protein sequence constraints. Such constraints appear as residues co-conserved in functionally related subgroups, as subtle pairwise correlations (i.e., direct couplings), and as correlations among these sequence features or with structural features. SPARC performs three types of analyses. First, based on pairwise sequence correlations, it estimates the biological relevance of alternative conformations and of homomeric contacts, as illustrated here for death domains. Second, it estimates the statistical significance of the correspondence between directly coupled residue pairs and interactions at heterodimeric interfaces. Third, given molecular dynamics simulated structures, it characterizes interactions among constrained residues or between such residues and ligands that: (a) are stably maintained during the simulation; (b) undergo correlated formation and/or disruption of interactions with other constrained residues; or (c) switch between alternative interactions. We illustrate this for two homohexameric complexes: the bacterial enhancer binding protein (bEBP) NtrC1, which activates transcription by remodeling RNA polymerase (RNAP) containing σ 54 , and for DnaB helicase, which opens DNA at the bacterial replication fork. Based on the NtrC1 analysis, we hypothesize possible mechanisms for inhibiting ATP hydrolysis until ADP is released from an adjacent subunit and for coupling ATP hydrolysis to restructuring of σ 54 binding loops. Based on the DnaB analysis, we hypothesize that DnaB ‘grabs’ ssDNA by flipping every fourth base and inserting it into cavities between subunits and that flipping of a DnaB-specific glutamine residue triggers ATP hydrolysis.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Duration of super-emitting oil and gas methane sources

The duration of super-emitting events (>100 kg h -1 ) in oil and gas basins remains insufficiently understood but is key for reporting programs and mitigation strategies. Carbon Mapper conducted aerial surveys from April 30 to May 17, 2024, over the New Mexico Permian Basin, covering 276,000 wells, 1100 compressor stations, 175 gas processing plants, and 27,000 km of pipeline. We find over 500 super-emitting sources with 300 of these sources observed repeatedly across multiple days. We quantify total super emissions by integrating individual events with observationally constrained event durations (5.98 −14.7 Gg CH 4 ) and compare to total emissions derived from basin average snapshots (12.7 ± 0.92 Gg CH 4 ). This gap between emission estimates is reconciled through assumptions on missed detections, characteristic event duration, detection frequency, and diurnal variability. Emission events generally lasted for at least 2 hours, and a small subset of sources (18 total), persistently emitted throughout the entire campaign, representing a near-term opportunity for mitigation. When compared to regional flux estimates derived from independent observations, we estimate super-emitters to contribute approximately 50% (37-73%) towards total emissions. Frequent wide-area monitoring is crucial for capturing rare super-emitter events that, together with other emission sources, drive basin-level variability and emission intensity.

Cusworth, Daniel H. [Carbon Mapper, Pasadena, CA (↗
DOE OSTI2025

Search for flavor-changing neutral current interactions of the top quark mediated by a Higgs boson in proton-proton collisions at 13 TeV

A search for flavor-changing neutral current interactions of the top quark (𝑡) and the Higgs boson (𝐻) is presented. The search is based on proton-proton collision data collected in 2016–2018 at a center-of-mass energy of 13 TeV with the CMS detector at the LHC, and corresponding to an integrated luminosity of 138 fb −1 . Events containing a pair of leptons with the same-sign electric charge and at least one jet are considered. The results are used to constrain the branching fraction (ℬ) of the top quark decaying to a Higgs boson and an up (𝑢) or charm (𝑐) quark. No significant excess above the estimated background was found. The observed (expected) upper limits at a 95% confidence level are found to be 0.072% (0.059%) for ℬ⁡(𝑡 → 𝐻⁡𝑢) and 0.043% (0.062%) for ℬ⁡(𝑡 → 𝐻⁡𝑐). These results are combined with two other searches performed by the CMS Collaboration for flavor-changing neutral current interactions of top quarks and Higgs bosons in final states where the Higgs boson decays to either a pair of photons or a pair of bottom quarks. The resulting observed (expected) upper limits at the 95% confidence level are 0.019% (0.027%) for ℬ⁡(𝑡 → 𝐻⁡𝑢) and 0.037% (0.035%) for ℬ⁡(𝑡 → 𝐻⁡𝑐).

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Atacama Cosmology Telescope: Multiprobe cosmology with unWISE galaxies and ACT DR6 CMB lensing

We present a joint analysis of the cosmic microwave background (CMB) lensing power spectra measured from the Data Release 6 of the Atacama Cosmology Telescope (ACT) and Planck PR4, cross-correlations between the ACT and Planck lensing reconstruction and galaxy clustering from unWISE, and the unWISE clustering auto-spectrum. We obtain 1.5% constraints on the matter density fluctuations at late times parametrized by the best constrained parameter combination 𝑆$^{3⁢x⁢2⁢pt}_{8}$ ≡ 𝜎 8 ⁢(Ω 𝑚 /0.3) 0.4 = 0.815 ± 0.012. The commonly used 𝑆 8 ≡ 𝜎 8⁢ (Ω 𝑚 /0.3) 0.5 parameter is constrained to 𝑆 8 = 0.816 ± 0.015. In combination with baryon acoustic oscillation (BAO) measurements we find 𝜎 8 = 0.815 ± 0.012. We also present sound-horizon-independent estimates of the present day Hubble rate of 𝐻 0 = 66.4$^{+3.2}_{−3.7}$ km s −1 Mpc −1 from our large scale structure data alone and 𝐻 0 = 64.3$^{+2.1}_{−2.4}$ km s −1 Mpc −1 in combination with uncalibrated supernovae from Pantheon+. Using parametric estimates of the evolution of matter density fluctuations, we place constraints on cosmic structure in a range of high redshifts typically inaccessible with cross-correlation analyses. Combining lensing cross- and autocorrelations, we derive a 3.3% constraint on the integrated matter density fluctuations above 𝑧 = 2.4, one of the tightest constraints in this redshift range and fully consistent with a Λ cold dark matter (Λ⁢CDM) model fit to the primary CMB from Planck. Finally, combining with primary CMB observations and using the extended low redshift coverage of these combined datasets we derive constraints on a variety of extensions to the Λ⁢CDM model including massive neutrinos, spatial curvature, and dark energy. We find in flat Λ⁢CDM⁢ ∑𝑚 𝜈 < 0.12 eV at 95% confidence using the large scale structure data, BAO measurements from Sloan Digital Sky Survey, and primary CMB observations.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Improving the Parameterization of Cloud and Rain Microphysics in E3SM using Novel Observationally-Constrained Bayesian Approach (Final Technical Report)

In this project, we sought to develop new cloud and rain microphysics frameworks within the Energy Exascale Earth System Model (E3SM). This work encompassed two primary avenues of research: 1) Further development of a Bayesian-based scheme called BOSS (Bayesian Observationally-constrained Statistical-physical Scheme) to represent cloud and rain microphysics, testing it in realistic high-resolution cloud models, and implementing it in E3SM; 2) Development of a methodology utilizing machine learning to enable computationally tractable use of tractable use of Markov chain Monte Carlo sampling for Bayesian parameter estimation in Earth system and cloud models. In this project, we adapted the BOSS microphysics scheme, originally formulated for rain-only, to include all liquid-phase microphysical processes for cloud and rain, in particular the processes that mediate between these two categories, for example the conversion from cloud to rain through collision and coalescence of drops. We constrained the scheme via comparison and testing against a detailed model that explicitly represents the evolution of cloud and rain particles, called a bin microphysics scheme.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Extremely rapid, Lagrangian modeling of 2D flooding: A rivulet-based approach

Estimates of potential flood inundation areas and depths are critical to informing the preparedness, response, and investment decisions of many government agencies and private sector organizations, especially under a changing climate. The standard modeling approaches, however, are often either computationally intensive or constrained in their accuracy or applicability. A novel, rivulet-based, 2D model of pluvial flooding is described in this article that is 10,000 to 10 million times less computationally complex than the full solution of the shallow water equations, yet achieves inundation area hit rates of between 0.8 and 0.9 and relative absolute mean errors of 10%-20% across a wide range of flow depths. This combination of accuracy and efficiency will enable real-time depth estimates during flood events, detailed sensitivity analyses, and the generation of large ensembles to support broad uncertainty analyses.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Observational Constraints on the Cloud Feedback Pattern Effect

Abstract Model evidence for the “pattern effect” assumes that global climate models (GCMs) faithfully simulate how clouds respond to varying sea surface temperature (SST) patterns and associated meteorological perturbations. We exploit time-invariant satellite-based estimates of the sensitivity of marine low clouds to meteorological perturbations to estimate how these clouds responded to time-varying SST patterns and meteorology between 1870 and 2014. GCMs and reanalyses provide estimates of the historical meteorological changes. Observations suggest that increasing estimated inversion strength (EIS) between 1980 and 2014 produced a negative low cloud feedback, opposite to the positive feedback expected from increasing CO 2 . This indicates that the processes responsible for marine cloud changes from 1980 to the near present are distinct from those associated with an increase in CO 2 . We also observationally constrain the difference between the historical near-global marine low cloud feedback, , and that arising from increasing CO 2 , . We find that this cloud feedback pattern effect depends strongly on time period and reanalysis dataset, and that varying changes in EIS and SST with warming explain much of its variability. Between 1980 and 2014, we estimate that (90% confidence) assuming meteorological changes from the Multiple Reanalysis Ensemble, implying a total pattern effect (that arising from all climate feedbacks) of 1.86 ± 0.45 W m −2 K −1 . This observational evidence corroborates previous quantitative estimates of the pattern effect, which heretofore relied largely upon GCM-based cloud changes. However, disparate historical meteorological changes across individual reanalyses contribute to considerable uncertainty in its magnitude.

58 GEOSCIENCES↗
DOE OSTI2022

Late-onset Circumstellar Medium Interactions are Rare: An Unbiased GALEX View of Type Ia Supernovae

Abstract Using ultraviolet (UV) light curves, we constrain the circumstellar environments of 1080 Type Ia supernovae (SNe Ia) within z < 0.5 from archival Galaxy Evolution Explorer (GALEX) observations. All SNe Ia are required to have pre- and post-explosion GALEX observations to ensure adequate subtraction of the host-galaxy flux. Using the late-time GALEX observations, we look for the UV excess expected from any interaction between the SN ejecta and circumstellar material (CSM). Four SNe Ia are detected near maximum light, and we compare the GALEX photometry to archival data. However, we find that none of our targets show convincing evidence of CSM interaction. A recent Hubble Space Telescope (HST) survey estimates that ∼6% of SNe Ia may interact with distant CSM, but statistical inferences are complicated by the small sample size and selection effects. By injecting model light curves into our data and then recovering them, we constrain a broad range of CSM interactions based on the CSM interaction start time and the maximum luminosity. Combining our GALEX nondetections with the HST results, we constrain occurrence of late-onset CSM interaction among SNe Ia with moderate CSM interaction, similar to that observed in PTF11kx, to f CSM ≲ 5.1% between 0 and 500 days after discovery and ≲2.7% between 500 and 1000 days after discovery at 90% confidence. For weaker CSM interactions similar to SN 2015cp, we obtain limits of ≲16% and ≲4.8%, respectively, for the same time ranges.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

A comparison of Gaussian processes and neural networks for computer model emulation and calibration

The Department of Energy relies on complex physics simulations for prediction in domains like cosmology, nuclear theory, and materials science. These simulations are often extremely computationally intensive, with some requiring days or weeks for a single simulation. In order to assure their accuracy, these models are calibrated against observational data in order to estimate inputs and systematic biases. Because of their great computational complexity, this process typically requires the construction of an emulator, a fast approximation to the simulation. In this paper, two emulator approaches are compared: Gaussian process regression and neural networks. Their emulation accuracy and calibration performance on three real problems of Department of Energy interest is considered. On these problems, the Gaussian process emulator tends to be more accurate with narrower, but still well-calibrated uncertainty estimates. The neural network emulator is accurate, but tends to have large uncertainty on its predictions. Finally, as a result, calibration with the Gaussian process emulator produces more constrained posteriors that still perform well in prediction.

97 MATHEMATICS AND COMPUTING↗