Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “distributed inference”

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 235 records · Page 13
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Representing Complex Systems as Graphs for Debugging and Predictive Maintenance-Preliminary Thoughts

Representing complex systems as graphs enables use of mathematical tools to identify faults or predict failures. Graph nodes correspond to individual modules or subsystems, and edges link coupled system parts. ‘Probes’ measure the node outputs, monitoring the system health for unexpected behavior. Assuming one cannot probe every point, within a system, the fault correlates to a region—not necessarily the specific location. Bayesian networks trained to understand fault patterns can accurately identify the source. The diagnostic tool described aides debugging by pinpointing system failure causes. For predictive maintenance, probe data develop probability distribution functions describing subsystem mean time to failure. Unit lifetime can be estimated through these probability distributions. Two approaches include using Bayesian classifiers to infer the system failure source and developing maintenance schedules by treating systems as collections of random variables. When failure behavior does not follow a closed form function, use of similarity models is proposed.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Analysis of Spectral Lines in Large Databases of Synthetic Spectra for Massive Stars

In this paper, we describe a program that identifies in the optical spectrum the main parameters of a spectral line, namely the initial and final wavelengths, and the line depth. Moreover, using numerical calculations, it identifies and removes adjacent lines. Next, the program calculates the equivalent width and the FWHM. The software was tested in a sample of 300 lines in two databases of synthetic spectra generated by the CMFGEN and PoWR codes, and 300 lines in observed spectra from the IACOB database, showing a Gaussian distribution of relative errors, from which it is inferred that 80% of the measured lines have errors less than 17% and only 5% of the lines have errors greater than 26%. The program was also run on the entire database of 45,000 CMFGEN and 202 POWR synthetic spectra, generating a library of H i, He i, and He ii lines necessary to feed the FITspec code for the derivation of stellar parameters: effective temperature, surface gravity, and luminosity.

47 OTHER INSTRUMENTATION↗
DOE OSTI2020

Constraining the Milky Way Mass Profile with Phase-space Distribution of Satellite Galaxies

We estimate the Milky Way (MW) halo properties using satellite kinematic data including the latest measurements from Gaia DR2. With a simulation-based 6D phase-space distribution function (DF) of satellite kinematics, we can infer halo properties efficiently and without bias, and handle the selection function and measurement errors rigorously in the Bayesian framework. Applying our DF from the EAGLE simulation to 28 satellites, we obtain an MW halo mass of $M={1.23}_{-0.18}^{+0.21}\times {10}^{12}{M}_{\odot }$ and a concentration of $c={9.4}_{-2.1}^{+2.8}$ with the prior based on the M–c relation. The inferred mass profile is consistent with previous measurements but with better precision and reliability due to the improved methodology and data. Potential improvement is illustrated by combining satellite data and stellar rotation curves. Using our EAGLE DF and best-fit MW potential, we provide much more precise estimates of the kinematics for those satellites with uncertain measurements. Compared to the EAGLE DF, which matches the observed satellite kinematics very well, the DF from the semi-analytical model based on the dark-matter-only simulation Millennium II (SAM-MII) over-represents satellites with small radii and velocities. We attribute this difference to less disruption of satellites with small pericenter distances in the SAM-MII simulation. Finally, by varying the disruption rate of such satellites in this simulation, we estimate a ~5% scatter in the inferred MW halo mass among hydrodynamics-based simulations.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Multiwavelength Observations of a New Redback Millisecond Pulsar 4FGL J1910.7-5320

We present the study of multiwavelength observations of an unidentified Fermi Large Area Telescope (LAT) source, 4FGL J1910.7-5320, a new candidate redback millisecond pulsar binary. In the 4FGL 95% error region of 4FGL J1910.7-5320, we find a possible binary with a 8.36 hr orbital period from the Catalina Real-Time Transient Survey, confirmed by optical spectroscopy using the SOAR telescope. This optical source was recently independently discovered as a redback pulsar by the TRAPUM project, confirming our prediction. We fit the optical spectral energy distributions of 4FGL J1910.7-5320 with a blackbody model, inferring a maximum distance of 4.1 kpc by assuming that the companion fills its Roche lobe with a radius of R = 0.7 R ⊙ . Using a 12.6 ks Chandra X-ray observation, we identified an X-ray counterpart for 4FGL J1910.7-5320, with a spectrum that can be described by an absorbed power law with a photon index of 1.0 ± 0.4. The spectrally hard X-ray emission shows tentative evidence for orbital variability. Using more than 12 yr of Fermi-LAT data, we refined the position of the γ-ray source, and the optical candidate still lies within the 68% positional error circle. In addition to 4FGL J1910.7-5320, we find a variable optical source with a periodic signal of 4.28 hr inside the 4FGL catalog 95% error region of another unidentified Fermi source, 4FGL J2029.5-4237. However, the γ-ray source does not have a significant X-ray counterpart in an 11.7 ks Chandra observation, with a 3σ flux upper limit of 2.4 × 10 -14 erg cm -2 s -1 (0.3–7 keV). Moreover, the optical source is outside our updated Fermi-LAT 95% error circle. These observational facts all suggest that this new redback millisecond pulsar powers the gamma-ray source 4FGL J1910.7-5320 while 4FGL J2029.5-4237 is unlikely the γ-ray counterpart to the 4.28 hr variable.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Benchmarking of Different Inverse Point Kinetics Implementations for an Autocorrected Reactimeter Algorithm

In November 2017, the Transient Reactor Test Facility returned to operation. Since that time, many transient test series have been completed; each has provided valuable data for materials performance, reactor safety that can be applied in future designs. During each experimental series, detector count rates provided important information on the core behavior during transients. However, a limitation of these data is that variations in the neutron distribution during experiments cause errors when attempting to infer reactivity evolution from detector signals. Neutron physics codes can be used to compute the flux shape variations. However, this is a poor solution when the experimental data is used to do verification, validation and uncertainty quantification (VVUQ) on codes. Indeed, if the output of the code is used both as a reference and to correct what the reference is compared to, the circular dependency limits the quality of the VVUQ approach. To overcome this problem, an Autocorrected Reactimeter Algorithm (ACRA) has been developed. This approach infers time-dependent reactivity evolution by testing different spatial corrections and selecting the one that minimizes reactivity variations when the core is in a frozen configuration (i.e. when there is no variation in parameters affecting reactivity). However, the scope of this method was limited to transients where there were negligible thermal feedback. Indeed, the core is never in a frozen configuration when the fuel temperature varies during the whole transient. This is our motivation for the development of an improved version of the ACRA which does not require frozen configurations

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2020

A Deep Generative Model for Non-Intrusive Identification of EV Charging Profiles

The proliferation of electric vehicles (EVs) brings environmental benefits and technical challenges to power grids. An identification algorithm which can accurately extract individual EV charging profiles out of widely available smart meter measurements has attracted great interests. This paper proposes a non-intrusive identification framework for EV charging profile extraction, which is driven by deep generative models (DGM). First, the proposed DGM is designed as a representation layer embedded into the Markov process and used to model the joint probability distribution of available time-series data. A novel contribution is to approximate posterior distributions by neural networks whose parameters are obtained by variational inference and supervised learning. Second, the EV charging status is inferred from the DGM via dynamic programming. Lastly, the desired EV charging profile can be reconstructed by the rated power of EV models and inferred status. Compared with the benchmark Hidden Markov Models, the proposed framework can better handle noise in data with less computational complexity and better overall accuracy performances with smaller recall. The proposed framework is validated by numerical experiments on the Pecan Street dataset.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2024

The Effect of the Velocity Distribution on Kilonova Emission

The electromagnetic emission from the nonrelativistic ejecta launched in neutron star mergers (either dynamically or through a disk wind) has the potential to probe both the total mass and composition of this ejecta. These observations are crucial in understanding the role of these mergers in the production of r-process elements in the Universe. However, many properties of the ejecta can alter the light curves and we must both identify which properties play a role in shaping this emission and understand the effects these properties have on the emission before we can use observations to place strong constraints on the amount of r-process elements produced in the merger. This paper focuses on understanding the effect of the velocity distribution (amount of mass moving at different velocities) for lanthanide-rich ejecta on the light curves and spectra. The simulations use distributions guided by recent calculations of disk outflows and compare the velocity-distribution effects to those of ejecta mass, velocity, and composition. Our comparisons show that uncertainties in the velocity distribution can lead to a factor of 2–4 uncertainties in the inferred ejecta mass based on peak infrared luminosities. We also show that early-time UV or optical observations may be able to constrain the velocity distribution, reducing the uncertainty in the ejecta mass.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

The Effect of the Velocity Distribution on Kilonova Emission

Abstract The electromagnetic emission from the nonrelativistic ejecta launched in neutron star mergers (either dynamically or through a disk wind) has the potential to probe both the total mass and composition of this ejecta. These observations are crucial in understanding the role of these mergers in the production of r -process elements in the Universe. However, many properties of the ejecta can alter the light curves and we must both identify which properties play a role in shaping this emission and understand the effects these properties have on the emission before we can use observations to place strong constraints on the amount of r -process elements produced in the merger. This paper focuses on understanding the effect of the velocity distribution (amount of mass moving at different velocities) for lanthanide-rich ejecta on the light curves and spectra. The simulations use distributions guided by recent calculations of disk outflows and compare the velocity-distribution effects to those of ejecta mass, velocity, and composition. Our comparisons show that uncertainties in the velocity distribution can lead to a factor of 2–4 uncertainties in the inferred ejecta mass based on peak infrared luminosities. We also show that early-time UV or optical observations may be able to constrain the velocity distribution, reducing the uncertainty in the ejecta mass.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Risk Assessment in a Chemical Laboratory Following an Explosive Incident Involving a Novel Diazonium Compound: Retrospective Analysis and Lessons Learned

Diazonium compounds are synthetically useful in the production of dyes and textiles, however they are highly explosive under dry conditions. Explosion prevention becomes more difficult when new diazonium compounds are synthesized, because while some syntheses include a counterion to increase their stability, this is not always a reliable method to prevent an explosive incident. Due to the uncertainty surrounding the explosiveness of different diazonium compounds, it is important to understand how to safely clean up after an incident and how to determine when it is safe to return a laboratory to typical operational use, particularly when the incident involves a novel compound where a standard does not exist for instrument calibration. Here, an explosive event is discussed involving the synthesis of 4-bromo-benzenediazonium-2-carboxylate. Following the explosive incident and 3-step cleanup, which involved a precautionary neutralization step, samples were collected from the fume hood where the incident occurred. Because the incident involved an unstable, novel compound that is not commercially available and was deemed unsafe to resynthesize for instrument calibration, we assessed the risk of further explosion by analyzing for the stable decomposition products. Mass spectrometry analysis confirmed that the residue in the fume hood contained 5-bromosalicylic acid, a decomposition product of 4-bromo-benzenediazonium-2-carboxylate. Samples were taken from multiple points in the fume hood and analyzed to estimate the spatial distribution of the decomposition product. Based on this analysis, we inferred that the primary decomposition product was far more abundant than residual energetic, indicating the energetic had been consumed or neutralized to a trace quantity where the risk of further explosion was low. Furthermore, the steps presented here─specifically, initial neutralization and then analyzing the spatial distribution of expected decomposition products to assess risk when a novel explosive material is detonated in a confined space─were our approach to assess further risk following an explosion due to a novel diazonium compound without the need for any further handling or resynthesis of the energetic. Here, we present our approach and critically analyze these steps by discussing retrospective lessons learned and alternative analytical approaches.

Computer simulations↗
DOE OSTI2025

γIn-beam angular distribution and linear polarization measurements with GRETINA using a simple energy-ordering approach

Angular distribution and linear polarization measurements are powerful tools for inferring the spins and parities of nuclear levels. In this work, the performance of the Gamma-Ray Energy Tracking In-beam Nuclear Array (GRETINA) as a Compton polarimeter was characterized in a fusion-evaporation reaction experiment using a simple energy-ordering approach for the interaction points assigned in the signal decomposition process. A variety of multipolarities and characters for γ-ray transitions in the reaction products 25 Mg, 25 Na, and 22 Ne, formed from fusion-evaporation of an 18 O beam on a 9 Be target, were examined. The experimental angular distributions and linear polarization asymmetries were consistent with predictions using the theoretical formalism accounting for the Lorentz boost.

Angular distribution↗
DOE OSTI2023

Measuring the giant radio galaxy length distribution with the LoTSS

Many massive galaxies launch jets from the accretion disk of their central black hole, but only ~10 3 instances are known in which the associated outflows form giant radio galaxies (GRGs, or giants): luminous structures of megaparsec extent that consist of atomic nuclei, relativistic electrons, and magnetic fields. Large samples are imperative to understanding the enigmatic growth of giants, and recent systematic searches in homogeneous surveys constitute a promising development. For the first time, it is possible to perform meaningful precision statistics with GRG lengths, but a framework to do so is missing. We measured the intrinsic GRG length distribution by combining a novel statistical framework with a LOFAR Two-metre Sky Survey (LoTSS) sample of freshly discovered giants. In turn, this allowed us to answer an array of questions on giants. For example, we can now assess how rare a 5 Mpc giant is compared with one of 1 Mpc, and how much larger – given a projected length – the corresponding intrinsic length is expected to be. Notably, we can now also infer the GRG number density in the Local Universe. We assumed the intrinsic GRG length distribution to be Paretian (i.e. of power-law form) with tail index ξ, and predicted the observed distribution by modelling projection and selection effects. To infer ξ, we also systematically searched the LoTSS for hitherto unknown giants and compiled the largest catalogue of giants to date. We show that if intrinsic GRG lengths are Pareto distributed with index ξ, then projected GRG lengths are also Pareto distributed with index ξ. Selection effects induce curvature in the observed projected GRG length distribution: angular length selection flattens it towards the lower end, while surface brightness selection steepens it towards the higher end. We explicitly derived a GRG’s posterior over intrinsic lengths given its projected length, laying bare the ξ dependence. We also discovered 2060 giants within LoTSS DR2 pipeline products; our sample more than doubles the known population. Spectacular discoveries include the largest, second-largest, and fourth-largest GRG known (l p = 5.1 Mpc, l p = 5.0 Mpc, and l p = 4.8 Mpc), the largest GRG known hosted by a spiral galaxy (l p = 2.5 Mpc), and the largest secure GRG known beyond redshift 1 (l p = 3.9 Mpc). We increase the number of known giants whose angular length exceeds that of the Moon from 10 to 23; among the discoveries is the angularly largest known radio galaxy in the Northern Sky, which is also the angularly largest known GRG ($\phi$ = 2°). Combining theory and data, we determined that intrinsic GRG lengths are well described by a Pareto distribution, and measured the index ξ = –3.5 ± 0.5. This implies that, given its projected length, a GRG’s intrinsic length is expected to be just 15% larger. Finally, we determined the comoving number density of giants in the Local Universe to be n GRG = 5 ± 2(100 Mpc) –3 . We developed a practical mathematical framework that elucidates the statistics of giant radio galaxy lengths. Through a LoTSS search, we also discovered 2060 new giants. By combining both advances, we determined that intrinsic GRG lengths are well described by a Pareto distribution with index ξ = –3.5 ± 0.5, and that giants are truly rare in a cosmological sense: most clusters and filaments of the Cosmic Web are not currently home to a giant. Thus, our work yields new observational constraints for analytical models and simulations featuring radio galaxy growth.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Probabilistic Physics-Informed Graph Convolutional Network for Active Distribution System Voltage Prediction

Here this letter proposes a novel data-driven probabilistic physics-informed graph convolutional network (GCN) for active distribution system voltage prediction with PVs and EVs. It leverages both measurements and network topology to accurately and efficiently predict node voltages without the need for an accurate distribution system power flow model. The dropout-enabled Bayesian inference is developed to achieve uncertainty quantification of the voltage prediction. Thanks to the network model embedding, it also has robustness against topology changes, a key difference with existing machine learning-based approaches. Comparison results with other state-of-the-art machine learning methods on a realistic 759-node distribution system demonstrate that the proposed method can achieve better accuracy and robustness under different scenarios.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2021

Integration of a DER Management System in Riverside. Final report

The tasks in this project covered various aspects, including algorithm development, algorithm integration into a commercial Active Network Management (ANM) platform, hardware-in-the-loop (HIL) testing in an industry-standard testing platform, pilot demonstration in Riverside, California, and also cost and benefit analysis. The DERMS platform in this project can host different algorithms developed on different platforms (e.g., MATLAB and Python) and it can interact with different hardware devices (e.g., different PV inverters, battery inverters, and different sensors). The DER control solution are based on an advanced model-free, layered, and clustered DER control paradigm. At the core of the DER control algorithms was the concept of Extremum Seeking (ES), which is a model-free probing-based control technique. The ES-based control algorithms were tested on major real-world inverters; both individually and in a cluster. It was shown that even legacy equipment (or when paired with a few additional advanced equipment) can support such advanced control. The monitoring algorithms utilize a heterogeneous set of legacy and advanced sensor measurements, such as behind-the-meter DER sensors, distribution-level Phase Measurement Units, distribution-substation Supervisory Control and Data Acquisition (SCADA), and line current sensors, with their limited availability; in order to infer practical network conditions. Sensor data are utilized to achieve resource forecasting, phase identification, and distribution system state estimation. The technology that was developed and demonstrated in this project could be transformational to utilities, including the smaller municipal utilities such as in Riverside, which may not have the resources to deploy advanced distribution system and DERMS solutions in order to support high penetration of solar power integration. This project created a real-world prototype to provide utilities with an assessment of smart grid monitoring and control technologies.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2025

Low-Jitter Clock Receivers for Fast Timing Applications

Precision timing is a key requirement for emerging 4D particle tracking, Positron Emission Tomography (PET), beam and fusion plasma diagnostics, and other systems. Time-to-Digital Converters (TDCs) are commonly used to provide digital estimates of the relative timing between events, but the jitter performance of a TDC can be no better than the performance of the circuits that acquire the pulses and deliver them to the TDC. Several clock receiver and distribution circuits were evaluated, and a differential amplifier with resistive loads driving a pseudo-differential clock distribution network, developed using design guidelines for radiation tolerance and cryogenic compatibility, was fabricated as part of three prototypes: an analog front-end testbed chip for high-precision timing pixel readout, a dedicated TDC evaluation chip, and a Low-Gain Avalanche Detector (LGAD) readout circuit. Based on TDC measurements of the prototypes, we infer that the jitter added by the clock receiver and distribution circuits is less than 2.25 ps-rms. This performance meets the requirements of many future precision timing systems. The clock receiver and on-chip pseudo-differential driver were fabricated in commercial 28-nm CMOS technology and occupy 2288 µm 2 .

47 OTHER INSTRUMENTATION↗
DOE OSTI2022

Generative Ensemble Regression: Learning Particle Dynamics from Observations of Ensembles with Physics-informed Deep Generative Models

Here, we propose a new method for inferring the governing stochastic ordinary differential equations (SODEs) by observing particle ensembles at discrete and sparse time instants, i.e., multiple “snapshots.” Particle coordinates at a single time instant, possibly noisy or truncated, are recorded in each snapshot but are unpaired across the snapshots. By training a physics-informed generative model that generates “fake” sample paths, we aim to fit the observed particle ensemble distributions with a curve in the probability measure space, which is induced from the inferred particle dynamics. We employ different metrics to quantify the differences between distributions, e.g., the sliced Wasserstein distances and the adversarial losses in generative adversarial networks. We refer to this method as generative “ensemble-regression” (GER), in analogy to the classic “point-regression,” where we infer the dynamics by performing regression in the Euclidean space. We illustrate the GER by learning the drift and diffusion terms of particle ensembles governed by SODEs with Brownian motions and Lévy processes up to 100 dimensions. We also discuss how to treat cases with noisy or truncated observations. Apart from systems consisting of independent particles, we also tackle nonlocal interacting particle systems with unknown interaction potential parameters by constructing a physics-informed loss function. Finally, we investigate scenarios of paired observations and discuss how to reduce the dimensionality in such cases by proving a convergence theorem that provides theoretical support.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Imaging electron angular distributions to assess a full-power petawatt-class laser focus

Here we present a technique to assess the focal volume of petawatt-class lasers at full power. Our approach exploits quantitative measurement of the angular distribution of electrons born in the focus via ionization of rarefied gas, which are accelerated forward and ejected ponderomotively by the field. We show that a bivariate (θ,Φ) angular distribution, which was obtained with image plates, not only enables the peak intensity to be extracted, but also reflects nonideality of the focal-spot intensity distribution. In our prototype demonstration at intensities of a few ×10 19 to a few ×10 20 W/cm 2 , an f/10 optic produced a focal spot in the paraxial regime. This allows a plane-wave parametrization of the peak intensity given by tanθ c =2/a 0 (a 0 being the normalized vector potential and θ c the minimum ejection angle) to be compared with our measurements. Qualitative agreement was found using an a 0 inferred from the pulse energy, pulse duration, and focal spot distribution with a modified parametrization, tanθ c =2η/a 0 ($η=2.02^{+0.26}_{–0.22}$). This highlights the need for (i) better understanding of intensity degradation due to focal-spot distortions and (ii) more robust modeling of the ejection dynamics. Using single-shot detection of electrons, we showed that while there is significant shot-to-shot variation in the number of electrons ejected at a given angular position, the average distribution scales with the pulse energy in a way that is consistent with that seen with the image plates. Finally, we note that the asymptotic behavior as θ→0° limits the usability of angular measurement. For 800 nm, this limit is at an intensity ~10 21 W/cm 2 .

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Towards robust autonomous impedance spectroscopy analysis: A calibrated hierarchical Bayesian approach for electrochemical impedance spectroscopy (EIS) inversion

Distribution-based analyses, such as the distribution of relaxation times (DRT) and the distribution of diffusion times (DDT), present model-free alternatives to equivalent circuit modeling for analysis of electrochemical impedance spectroscopy (EIS) data. However, reconstructing such distributions from noisy impedance data is an ill-posed problem that must be solved with specialized inversion algorithms, requiring careful control and tuning. Furthermore, most inversion algorithms developed to date can only solve problems of limited complexity. Herein, we present a new hierarchical Bayesian method for EIS inversion, leveraging efficient algorithms for optimization and Hamiltonian Monte Carlo (HMC) sampling to solve models of arbitrary complexity. We overcome the challenge of ad-hoc parameter tuning by encoding intrinsic characteristics of the DRT and DDT into flexible prior distributions and “pre-calibrating” the model to simulated data. This approach is versatile, highly robust to noise, and provides quantitative estimates of both the error structure of the data and the uncertainty in the recovered distributions. The model is validated with simulated data to demonstrate accurate recovery of the DRT and the DDT. The method also shows promise for simultaneous recovery of multiple distributions, raising the intriguing possibility of semi-autonomous EIS analysis and ad-hoc model construction. Finally, the practical utility of the method is illustrated with experimental data. Throughout, we draw comparisons to several recently published EIS inversion methodologies.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Hierarchical Inference of the Lensing Convergence from Photometric Catalogs with Bayesian Graph Neural Networks

Abstract We present a Bayesian graph neural network (BGNN) that can estimate the weak lensing convergence ( κ ) from photometric measurements of galaxies along a given line of sight (LOS). The method is of particular interest in strong gravitational time-delay cosmography (TDC), where characterizing the “external convergence” ( κ ext ) from the lens environment and LOS is necessary for precise Hubble constant ( H 0 ) inference. Starting from a large-scale simulation with a κ resolution of ∼1′, we introduce fluctuations on galaxy–galaxy lensing scales of ∼1″ and extract random sight lines to train our BGNN. We then evaluate the model on test sets with varying degrees of overlap with the training distribution. For each test set of 1000 sight lines, the BGNN infers the individual κ posteriors, which we combine in a hierarchical Bayesian model to yield constraints on the hyperparameters governing the population. For a test field well sampled by the training set, the BGNN recovers the population mean of κ precisely and without bias (within the 2 σ credible interval), resulting in a contribution to the H 0 error budget well under 1%. In the tails of the training set with sparse samples, the BGNN, which can ingest all available information about each sight line, extracts a stronger κ signal compared to a simplified version of the traditional method based on matching galaxy number counts, which is limited by sample variance. Our hierarchical inference pipeline using BGNNs promises to improve the κ ext characterization for precision TDC. The code is available as a public Python package, Node to Joy ⏬ .

79 ASTRONOMY AND ASTROPHYSICS↗