Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “data amplification”

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 19 records
All sourcesNASA NTRSDOE OSTI
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↗
DOE OSTI2025

The Bias-Variance-Correlation Tradeoff and Its Implications for ML Applications in HEP

The bias-variance tradeoff is a well-recognized phenomenon in statistics and machine learning. In this talk, I will discuss an extension, dubbed the bias-variance-correlation tradeoff. Roughly speaking, as the flexibility of a model decreases, the correlations in the outputs of a trained model for different inputs increases. Such correlations have implications for several applications of machine learning in high energy physics, e.g., the use generative models for event generation. In particular, I will argue that claims in the literature of data amplification by generative models stem from ignoring important correlations between the model's outputs for different inputs.

Shyamsundar, Prasanth [Fermilab] (ORCID:0000000227↗
DOE OSTI2021

Revisiting matrix-based inversion of scanning mobility particle sizer (SMPS) and humidified tandem differential mobility analyzer (HTDMA) data

Abstract. Tikhonov regularization is a tool for reducing noise amplification during data inversion. This work introduces RegularizationTools.jl, a general-purpose software package for applying Tikhonov regularization to data. The package implements well-established numerical algorithms and is suitable for systems of up to ∼ 1000 equations. Included is an abstraction to systematically categorize specific inversion configurations and their associated hyperparameters. A generic interface translates arbitrary linear forward models defined by a computer function into the corresponding design matrix. This obviates the need to explicitly write out and discretize the Fredholm integral equation, thus facilitating fast prototyping of new regularization schemes associated with measurement techniques. Example applications include the inversion involving data from scanning mobility particle sizers (SMPSs) and humidified tandem differential mobility analyzers (HTDMAs). Inversion of SMPS size distributions reported in this work builds upon the freely available software DifferentialMobilityAnalyzers.jl. The speed of inversion is improved by a factor of ∼ 200, now requiring between 2 and 5 ms per SMPS scan when using 120 size bins. Previously reported occasional failure to converge to a valid solution is reduced by switching from the L-curve method to generalized cross-validation as the metric to search for the optimal regularization parameter. Higher-order inversions resulting in smooth, denoised reconstructions of size distributions are now included in DifferentialMobilityAnalyzers.jl. This work also demonstrates that an SMPS-style matrix-based inversion can be applied to find the growth factor frequency distribution from raw HTDMA data while also accounting for multiply charged particles. The outcome of the aerosol-related inversion methods is showcased by inverting multi-week SMPS and HTDMA datasets from ground-based observations, including SMPS data obtained at Bodega Marine Laboratory during the CalWater 2/ACAPEX campaign and co-located SMPS and HTDMA data collected at the US Department of Energy observatory located at the Southern Great Plains site in Oklahoma, USA. Results show that the proposed approaches are suitable for unsupervised, nonparametric inversion of large-scale datasets as well as inversion in real time during data acquisition on low-cost reduced-instruction-set architectures used in single-board computers. The included software implementation of Tikhonov regularization is freely available, general, and domain-independent and thus can be applied to many other inverse problems arising in atmospheric measurement techniques and beyond.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Studies of BONuS12 Radial GEM Detector and TCS Beam Spin Asymmetry in CLAS12

The Barely Offshell Nucleon Structure (BONuS12) experiment adopted the concept of spectator tagging technique to study the nearly-free neutron structure function Fn2 in the CLAS12 of Jefferson Lab. A novel Radial Time Projection Chamber (RTPC) detector was built, tested and integrated into the CLAS12 system to detect a back-moving low momentum tagged proton in d(e,ep)X deep-inelastic scattering. It was a 40 cm long gaseous detector consisting of 3 layers of cylindrical GEM foils for the charge amplification, with the data readout directly from the surrounding padboard. The RTPC detected the recoiling spectator proton, in coincidence with the scattered electron in the CLAS12. Nucleon structure functions are directly related to the partonic functions, quarks momentum distribution in one dimension. A Generalized Parton Distribution (GPD) came to the lime-light as it encodes the information of both longitudinal momentum and transverse position of partons inside the nucleons. Factorization of hard process such as DVCS allows to access GPDs. Timelike Compton Scattering (TCS), ?p ? ??p, is another process that allows to access the GPDs. TCS is studied experimentally in the CLAS12 of Jefferson lab using the quasi-real photoproduction of time-like photon which eventually decays to lepton pair. This dissertation presents the concept of spectator tagging in BONuS12, and the research and development efforts during the BONuS12 preparation leading up to the successful data-taking during spring and summer 2020. In addition, analysis framework to extract the beam spin asymmetry of TCS events through the CLAS12 Run group A data is presented.

Poudel, Jiwan↗
DOE OSTI2018

Viral Communities of Shark Bay Modern Stromatolites

Single stranded DNA viruses have been previously shown to populate the oceans on a global scale, and are endemic in microbialites of both marine and freshwater systems. We undertook for the first time direct viral metagenomic shotgun sequencing to explore the diversity of viruses in the modern stromatolites of Shark Bay Australia. The data indicate that Shark Bay marine stromatolites have similar diversity of ssDNA viruses to that of Highbourne Cay, Bahamas. ssDNA viruses in cluster uniquely in Shark Bay and Highbourne Cay, potentially due to enrichment by phi29-mediated amplification bias. Further, pyrosequencing data was assembled from the Shark Bay systems into two putative viral genomes that are related to Genomoviridae family of ssDNA viruses. In addition, the cellular fraction was shown to be enriched for antiviral defense genes including CRISPR-Cas, BREX (bacteriophage exclusion), and DISARM (defense island system associated with restriction-modification), a potentially novel finding for these systems. This is the first evidence for viruses in the Shark Bay stromatolites, and these viruses may play key roles in modulating microbial diversity as well as potentially impacting ecosystem function through infection and the recycling of key nutrients.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2019

Testing dark matter and modifications to gravity using local Milky Way observables

Galactic rotation curves are often considered the first robust evidence for the existence of dark matter. Yet, even in the presence of a dark matter halo, other galactic-scale observations, such as the Baryonic Tully-Fisher Relation and the Radial Acceleration Relation, remain challenging to explain. This has motivated long-distance, infrared modifications to gravity as an alternative to the dark matter hypothesis as well as various DM theories with similar phenomenology. In general, the standard lore has been that any model that reduces to the phenomenology of MOdified Newtonian Dynamics (MOND) on galactic scales explains essentially all galaxy-scale observables. We present a framework to test precisely this statement using local Milky Way observables, including the vertical acceleration field, the rotation curve, the baryonic surface density, and the stellar disk profile. We focus on models that predict scalar amplifications of gravity, i.e., models that increase the magnitude but do not change the direction of the gravitational acceleration. We find that models of this type are disfavored relative to a simple dark matter halo model because the Milky Way data necessitates a substantial amplification of the radial acceleration with little amplification of the vertical acceleration. We conclude that models which result in a MOND-like force struggle to simultaneously explain both the rotational velocity and vertical motion of nearby stars in the Milky Way.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Identifying Interaction Location in SuperCDMS Detectors

The Super Cryogenic Dark Matter Search (SuperCDMS) experiment uses silicon and germanium particle detectors operated at temperatures of ∼ 30 mK to search for Weakly Interacting Massive Particles (WIMPs), which are candidate dark matter particles that interact weakly with nuclei in the detectors. In operating these detectors, it is required not only to measure the energy of the interaction between the WIMP and the nuclei, but also to reconstruct where the interaction occurred, as the location can be used to separate background interactions from signal and to correct for variations with the location of the energy response. In this project, we, as a team from the University of Minnesota, aim to address the problem of accurately reconstructing the locations of interactions in the SuperCDMS detectors using machine learning methods. The dataset we provided here includes interactions at thirteen different locations from test data taken at the University of Minnesota. For each interaction, a set of parameters was extracted from the signals from each of the five sensors. These parameters represent information known to be sensitive to interaction location, including the relative timing between pulses in different channels, and features like the pulse shape. The relative amplitudes of the pulses are also relevant but due to instabilities in amplification during the test, this data is not included. The parameters included for each interaction are described in our project document. For more details, feel free to check our Github page: https://fair-umn.github.io/FAIR-UMN-CDMS/

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2018

Optimizing de novo genome assembly from PCR-amplified metagenomes

Background. Metagenomics has transformed our understanding of microbial diversity across ecosystems, with recent advances enabling de novo assembly of genomes from metagenomes. These metagenome-assembled genomes are critical to provide ecological, evolutionary, and metabolic context for all the microbes and viruses yet to be cultivated. Metagenomes can now be generated from nanogram to subnanogram amounts of DNA. However, these libraries require several rounds of PCR amplification before sequencing, and recent data suggest these typically yield smaller and more fragmented assemblies than regular metagenomes. Methods. Here we evaluate de novo assembly methods of 169 PCR-amplified metagenomes, including 25 for which an unamplified counterpart is available, to optimize specific assembly approaches for PCR-amplified libraries. We first evaluated coverage bias by mapping reads from PCR-amplified metagenomes onto reference contigs obtained from unamplified metagenomes of the same samples. Then, we compared different assembly pipelines in terms of assembly size (number of bp in contigs ≥ 10kb) and error rates to evaluate which are the best suited for PCR-amplified metagenomes. Results. Read mapping analyses revealed that the depth of coverage within individual genomes is significantly more uneven in PCR-amplified datasets versus unamplified metagenomes, with regions of high depth of coverage enriched in short inserts. This enrichment scales with the number of PCR cycles performed, and is presumably due to preferential amplification of short inserts. Standard assembly pipelines are confounded by this type of coverage unevenness, so we evaluated other assembly options to mitigate these issues. We found that a pipeline combining read deduplication and an assembly algorithm originally designed to recover genomes from libraries generated after whole genome amplification (single-cell SPAdes) frequently improved assembly of contigs ≥ 10kb by 10 to 100-fold for low input metagenomes. Conclusions. PCR-amplified metagenomes have enabled scientists to explore communities traditionally challenging to describe, including some with extremely low biomass or from which DNA is particularly difficult to extract. Here we show that a modified assembly pipeline can lead to an improved de novo genome assembly from PCR-amplified datasets, and enables a better genome recovery from low input metagenomes.

Roux, Simon↗
DOE OSTI2018

Optimizing de novo genome assembly from PCR-amplified metagenomes

Background: Metagenomics has transformed our understanding of microbial diversity across ecosystems, with recent advances enabling de novo assembly of genomes from metagenomes. These metagenome-assembled genomes are critical to provide ecological, evolutionary, and metabolic context for all the microbes and viruses yet to be cultivated. Metagenomes can now be generated from nanogram to subnanogram amounts of DNA. However, these libraries require several rounds of PCR amplification before sequencing, and recent data suggest these typically yield smaller and more fragmented assemblies than regular metagenomes. Methods: Here we evaluate de novo assembly methods of 169 PCR-amplified metagenomes, including 25 for which an unamplified counterpart is available, to optimize specific assembly approaches for PCR-amplified libraries. We first evaluated coverage bias by mapping reads from PCR-amplified metagenomes onto reference contigs obtained from unamplified metagenomes of the same samples. Then, we compared different assembly pipelines in terms of assembly size (number of bp in contigs ≥ 10kb) and error rates to evaluate which are the best suited for PCR-amplified metagenomes. Results: . Read mapping analyses revealed that the depth of coverage within individual genomes is significantly more uneven in PCR-amplified datasets versus unamplified metagenomes, with regions of high depth of coverage enriched in short inserts. This enrichment scales with the number of PCR cycles performed, and is presumably due to preferential amplification of short inserts. Standard assembly pipelines are confounded by this type of coverage unevenness, so we evaluated other assembly options to mitigate these issues. We found that a pipeline combining read deduplication and an assembly algorithm originally designed to recover genomes from libraries generated after whole genome amplification (single-cell SPAdes) frequently improved assembly of contigs ≥ 10kb by 10 to 100-fold for low input metagenomes. Conclusions: PCR-amplified metagenomes have enabled scientists to explore communities traditionally challenging to describe, including some with extremely low biomass or from which DNA is particularly difficult to extract. Here we show that a modified assembly pipeline can lead to an improved de novo genome assembly from PCR-amplified datasets, and enables a better genome recovery from low input metagenomes.

Roux, Simon↗
DOE OSTI2019

Optimizing de novo genome assembly from PCR-amplified metagenomes

Background. Metagenomics has transformed our understanding of microbial diversity across ecosystems, with recent advances enabling de novo assembly of genomes from metagenomes. These metagenome-assembled genomes are critical to provide ecological, evolutionary, and metabolic context for all the microbes and viruses yet to be cultivated. Metagenomes can now be generated from nanogram to subnanogram amounts of DNA. However, these libraries require several rounds of PCR amplification before sequencing, and recent data suggest these typically yield smaller and more fragmented assemblies than regular metagenomes. Methods. Here we evaluate de novo assembly methods of 169 PCR-amplified metagenomes, including 25 for which an unamplified counterpart is available, to optimize specific assembly approaches for PCR-amplified libraries. We first evaluated coverage bias by mapping reads from PCR-amplified metagenomes onto reference contigs obtained from unamplified metagenomes of the same samples. Then, we compared different assembly pipelines in terms of assembly size (number of bp in contigs ≥ 10 kb) and error rates to evaluate which are the best suited for PCR-amplified metagenomes. Results. Read mapping analyses revealed that the depth of coverage within individual genomes is significantly more uneven in PCR-amplified datasets versus unamplified metagenomes, with regions of high depth of coverage enriched in short inserts. This enrichment scales with the number of PCR cycles performed, and is presumably due to preferential amplification of short inserts. Standard assembly pipelines are confounded by this type of coverage unevenness, so we evaluated other assembly options to mitigate these issues. We found that a pipeline combining read deduplication and an assembly algorithm originally designed to recover genomes from libraries generated after whole genome amplification (single-cell SPAdes) frequently improved assembly of contigs ≥10 kb by 10 to 100-fold for low input metagenomes. Conclusions. PCR-amplified metagenomes have enabled scientists to explore communities traditionally challenging to describe, including some with extremely low biomass or from which DNA is particularly difficult to extract. Here we show that a modified assembly pipeline can lead to an improved de novo genome assembly from PCR-amplified datasets, and enables a better genome recovery from low input metagenomes.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

TRT Compton Scattering Data: Comparisons between LLNL and LANL: Pointwise

This memo is one of a pair of reports comparing Compton scattering data in the context of thermal radiation transport (TRT). This document focuses on Pointwise (PW) data while the other report focuses on Multigroup (MG) data. These complementary analyses, via PW comparisons, show if the same physics are used and if fundamental bugs are present, and, via MG comparisons, quantify macroscopic Compton effects such as mean amplification factors. No averaging of Compton data over photon energy is done for the PW comparisons, while lab-dependent numerical resolution choices become relevant for the MG comparisons. This document plots and quantifies differences in PW Compton scattering data between LANL and LLNL for four disparate temperatures on a logarithmically-spaced energy grid. PW comparisons means we use the received LLNL data as-is, without any averaging or integration over energy. In addition to LANL-to-LLNL comparisons, we investigate the impact of numerical choices on LANL PW data. For the purposes of this comparison, we neglect induced scattering.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

TRT Compton Scattering Data: Comparisons between LLNL and LANL: Multigroup

This memo is one of a pair of reports comparing Compton scattering data in the context of thermal radiation transport (TRT). This document focuses on Multigroup (MG) data while the other report focuses on Pointwise (PW) data. These complementary analyses, via PW comparisons, show if the same physics are used and if fundamental bugs are present, and, via MG comparisons, quantify macroscopic Compton effects such as mean amplification factors. No averaging of Compton data over photon energy is done for the PW comparisons, while lab-dependent numerical resolution choices become relevant for the MG comparisons. For this MG-comparison report, we make and document the numerical assumptions we use to convert LLNL PW data into MG data. This document plots and quantifies differences in Compton scattering data between MG LANL and LLNL for four disparate temperatures on a logarithmically-spaced photon-energy grid. In addition to LANL and LLNL data, we add a third and fourth independent vote from data generated by Brooks Kinch (LANL postdoc). For the purposes of these comparisons, we neglect induced scattering.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2020

Fluorescent amplification for next generation sequencing (FA-NGS) library preparation

BACKGROUND: Next generation sequencing (NGS) has become a universal practice in modern molecular biology. As the throughput of sequencing experiments increases, the preparation of conventional multiplexed libraries becomes more labor intensive. Conventional library preparation typically requires quality control (QC) testing for individual libraries such as amplification success evaluation and quantification, none of which occur until the end of the library preparation process. RESULTS: In this study, we address the need for a more streamlined high-throughput NGS workflow by tethering real-time quantitative PCR (qPCR) to conventional workflows to save time and implement single tube and single reagent QC. We modified two distinct library preparation workflows by replacing PCR and quantification with qPCR using SYBR Green I. qPCR enabled individual library quantification for pooling in a single tube without the need for additional reagents. Additionally, a melting curve analysis was implemented as an intermediate QC test to confirm successful amplification. Sequencing analysis showed comparable percent reads for each indexed library, demonstrating that pooling calculations based on qPCR allow for an even representation of sequencing reads. To aid the modified workflow, a software toolkit was developed and used to generate pooling instructions and analyze qPCR and melting curve data. CONCLUSIONS: We successfully applied fluorescent amplification for next generation sequencing (FA-NGS) library preparation to both plasmids and bacterial genomes. As a result of using qPCR for quantification and proceeding directly to library pooling, the modified library preparation workflow has fewer overall steps. Therefore, we speculate that the FA-NGS workflow has less risk of user error. The melting curve analysis provides the necessary QC test to identify and troubleshoot library failures prior to sequencing. While this study demonstrates the value of FA-NGS for plasmid or gDNA libraries, we speculate that its versatility could lead to successful application across other library types.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Regularized Differentiation for Bioburden Density Estimation in Planetary Protection

In this paper, we propose and investigate the performance of two novel shrinkage estimators for bioburden density estimation in planetary protection. The estimators are based on the regularized differentiation of a cumulative count of colony forming units collected throughout the data collecting session or the life cycle of the entire mission. The regularized differentiation recasts the problem of bioburden density estimation as a linear least squares problem. The least squares problem is then solved through regularization techniques, such as truncated singular value decomposition and penalized least squares. The regularization is necessary to avoid noise amplification during the differentiation of noisy data. The two regularization estimators are compared with four other commonly used estimators to simultaneously evaluate the means of multivariable independent Poisson distributions: the maximum likelihood, noninformative Bayes estimator with Jeffreys prior, Empirical Bayes using conjugate gamma-Poisson model with gamma parameters selected by method of moments, and the Clevenson-Zidek estimator. It is shown through computer-simulated data that the regularized differentiation based on ridge regression has the smallest mean-squared error among all estimators. The analysis of shrinkage mechanism implemented by regularized differentiation is performed, and it is shown that the regularized differentiation amounts to performing a weighted averaging of all the samples. The weights are determined by the regularization parameter automatically selected by the L-curve technique. Since the method of least squares makes no distributional assumptions about the data, it presents an attractive technique for bioburden density estimation when there are concerns about the misspecification of the distributional model. The paper concludes with the analysis of the bioburden data collected during InSight mission and directions for future work.

97 - MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Characterizing and mitigating coherent errors in a trapped ion quantum processor using hidden inverses

Quantum computing testbeds exhibit high-fidelity quantum control over small collections of qubits, enabling performance of precise, repeatable operations followed by measurements. Currently, these noisy intermediate-scale devices can support a sufficient number of sequential operations prior to decoherence such that near term algorithms can be performed with proximate accuracy (like chemical accuracy for quantum chemistry problems). While the results of these algorithms are imperfect, these imperfections can help bootstrap quantum computer testbed development. Demonstrations of these algorithms over the past few years, coupled with the idea that imperfect algorithm performance can be caused by several dominant noise sources in the quantum processor, which can be measured and calibrated during algorithm execution or in post-processing, has led to the use of noise mitigation to improve typical computational results. Conversely, benchmark algorithms coupled with noise mitigation can help diagnose the nature of the noise, whether systematic or purely random. Here, we outline the use of coherent noise mitigation techniques as a characterization tool in trapped-ion testbeds. We perform model-fitting of the noisy data to determine the noise source based on realistic physics focused noise models and demonstrate that systematic noise amplification coupled with error mitigation schemes provides useful data for noise model deduction. Further, in order to connect lower level noise model details with application specific performance of near term algorithms, we experimentally construct the loss landscape of a variational algorithm under various injected noise sources coupled with error mitigation techniques. This type of connection enables application-aware hardware codesign, in which the most important noise sources in specific applications, like quantum chemistry, become foci of improvement in subsequent hardware generations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2020

More chemical detection through less sampling: amplifying chemical signals in hyperspectral data cubes through compressive sensing

Compressive sensing (CS) is a method of sampling which permits some classes of signals to be reconstructed with high accuracy even when they were sampled at sub-Nyquist rates. In this paper we explore a phenomenon in which bandwise CS sampling of a hyperspectral data cube followed by reconstruction can actually result in amplification of chemical signals contained in the cube. Perhaps most surprisingly, chemical signal amplification generally seems to increase as the level of sampling decreases. In some examples, the chemical signal is significantly stronger in a data cube reconstructed from 10% CS sampling than it is in the raw, 100% sampled data cube. We explore this phenomenon in two real-world datasets including the Physical Sciences Inc. Fabry-Perot interferometer sensor multispectral dataset and the Johns Hopkins Applied Physics Lab FTIR-based longwave infrared sensor hyperspectral dataset. Each of these datasets contains the release of a chemical simulant, such as glacial acetic acid, triethyl phospate, and sulfur hexafluoride, and in all cases we use the adaptive coherence estimator (ACE) to detect a target signal in the hyperspectral data cube. We end the paper by suggesting some theoretical justifications for why chemical signals would be amplified in CS sampled and reconstructed hyperspectral data cubes and discuss some practical implications.

Hyperspectral imagery, Compressive Sensing, single↗
DOE OSTI2023

Correlations between asymmetric compression, burn amplification, and hot-spot velocities in inertial confinement fusion implosions

Here, this manuscript examines the correlations between the hot-spot velocity (an observable signature of residual kinetic energy), low-mode implosion asymmetries, and burn amplification in inertial confinement fusion implosions on the National Ignition Facility (NIF). Using a combination of two-dimensional axis-symmetric and three-dimensional radiation-hydrodynamic simulations coupled to neutronics, we find that for typical NIF implosions, the stagnation asymmetry multiplies the observed hot-spot velocity anywhere from 80% to 120%, while burn amplification always increases it. Additionally, we find stagnation asymmetry typically deflects the observed hot-spot flow. The two mechanisms (low-mode implosion asymmetries and burn amplification) can be decoupled, and application of a simple model to a database of cryogenic implosions on the NIF infers the total hot-spot velocity amplification. This finding modifies the interpretation of data collected from inertial confinement fusion experiments and impacts the magnitude and origin of low-mode asymmetries.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

$\mathrm{SageNet}$: Fast Neural Network Emulation of the Stiff-amplified Gravitational Waves from Inflation

Accurate modeling of the inflationary gravitational waves (GWs) requires time-consuming, iterative numerical integrations of differential equations to take into account their backreaction on the expansion history. To improve computational efficiency while preserving accuracy, we present the Stiff-amplified Gravitational-wave Emulator Network (SageNet), a deep learning framework designed to replace conventional numerical solvers (code available at https://github.com/YifangLuo/SageNet). SageNet employs a long short-term memory architecture to emulate the present-day energy density spectrum of the inflationary GWs with possible stiff amplification, Ω GW (f). Trained on a data set of 25,689 numerically generated solutions, SageNet allows accurate reconstructions of Ω GW (f) and generalizes well to a wide range of cosmological parameters; 90.9% of the test emulations with randomly distributed parameters exhibit errors of under 4%. In addition, SageNet demonstrates its ability to learn and reproduce the artificial, adaptive sampling patterns in numerical calculations, which implement denser sampling of frequencies around changes in spectral indices in Ω GW (f). The dual capability of learning both physical and artificial features of the numerical GW spectra establishes SageNet as a robust alternative to exact numerical methods. Finally, our benchmark tests show that SageNet reduces the computation time from tens of seconds to milliseconds, achieving a speedup of ∼10 4 times over standard CPU-based numerical solvers with the potential for further acceleration on GPU hardware. These capabilities make SageNet a powerful tool for accelerating Bayesian inference procedures for extended cosmological models. In a broad sense, the SageNet framework offers a fast, accurate, and generalizable solution to modeling cosmological observables whose theoretical predictions demand costly differential equation solvers.

Astronomy data modeling↗