Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “ensemble prediction”

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 343 records · Page 19
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Application of artificial intelligence methods in the international roughness index prediction of rigid and composite pavements: a systematic review

The International Roughness Index (IRI) is a widely adopted metric for quantifying pavement roughness, directly influencing vehicle safety, ride comfort, and overall roadway performance. In recent years, the use of Machine Learning (ML) models for IRI prediction has gained momentum, with the goal of improving the allocation of maintenance and rehabilitation resources by enabling accurate assessments of pavement conditions. Most prior reviews, however, have concentrated on flexible pavements, leaving a notable gap regarding rigid and composite pavements. To address this gap, the present study conducts a systematic review of Artificial Intelligence (AI) methods applied to IRI prediction for rigid and composite pavements. Literature published between 2004 and 2025 is synthesized to highlight prevailing trends, methodological contributions, and directions for future research. Particular attention is given to the types of models employed, the datasets used for training and validation, and the role of input variables and data-processing strategies. Across the included studies, ensemble learning methods (especially gradient boosting variants such as XGBoost), artificial neural networks, and hybrid architectures frequently achieved high predictive skill, with several models reporting test-set coefficients of determination approaching 0.9–0.96, indicating strong potential for capturing the influence of traffic, pavement structure, and climatic factors. Since these results are obtained from heterogeneous datasets and evaluation protocols, they are interpreted qualitatively rather than as strict cross-study rankings. Analysis of input variables revealed that pavement age and initial IRI were included in 91% (21 of 23) and 78% (18 of 23) of studies, respectively. Climatic variables such as the freezing index appeared in 57% (13 of 23), while traffic-related factors were considered in 65% (15 of 23). The findings underscore the importance of standardized, high-quality datasets, such as those from the Long-Term Pavement Performance (LTPP) program, along with data consistency, model interpretability, computational efficiency, and replicability in enhancing IRI prediction. Future research should focus on incorporating input variable selection techniques to identify the most influential predictors, thereby improving accuracy and robustness. Integrating these approaches with advanced non-linear data-driven models, coupled with robust hyperparameter optimization, holds considerable promise for strengthening the reliability of IRI prediction and supporting resilient pavement management strategies.

42 ENGINEERING↗
DOE OSTI2021

Learning electric vehicle driver range anxiety with an initial state of charge-oriented gradient boosting approach

This manuscript focuses on the modeling of electric vehicle (EV) driver’s range anxiety, a fear that a vehicle does not have sufficient range, or state of charge (SOC) of the battery pack, to reach its destination and would strand its occupants. Despite numerous research studies on the modeling of charging behaviors, modeling efforts to understand at what battery percentages do EV drivers charge their vehicles, and what are the associated contributing factors, are rather limited. To this end, an ensemble learning model based on gradient boosting is developed. The model sequentially fits new predictors to new residuals of the previous prediction and, then, minimizes the loss when adding the latest prediction. A total of 18 features are defined and extracted from the multisource data, which cover information on driver, vehicles, stations, traffic conditions, as well as spatial-temporal context information of the charging events. The analyzed dataset includes 4.5-year’s charging event log data from 3,096 users and 468 public charging stations in Kansas City Missouri, and the macroscopic travel demand model maintained by the metropolitan planning organization. Here, the result shows the proposed model achieved a satisfactory result with a R square value of 0.54 and root mean square error of 0.14, both better than multiple linear regression model and random forest model. To reduce range anxiety, it is suggested that the priorities of deploying new charging facilities should be given to the areas with higher daily traffic prediction, with more conservative EV users or that are further from residential areas.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2022

The Mira–Titan Universe – IV. High-precision power spectrum emulation

Modern cosmological surveys are delivering data sets characterized by unprecedented quality and statistical completeness; this trend is expected to continue in the future as new ground- and space-based surveys come online. In order to maximally extract cosmological information from these observations, matching theoretical predictions are needed. At low redshifts, the surveys probe the non-linear regime of structure formation where cosmological simulations are the primary means of obtaining the required information. The computational cost of sufficiently resolved large-volume simulations makes it prohibitive to run very large ensembles. Nevertheless, precision emulators built on a tractable number of high-quality simulations can be used to build very fast prediction schemes to enable a variety of cosmological inference studies. We have recently introduced the Mira–Titan Universe simulation suite designed to construct emulators for a range of cosmological probes. This gravity-only set of simulations covers the standard six cosmological parameters {ω m , ω b , σ 8 , $h, n_s, w_0$} and, in addition, includes massive neutrinos and a dynamical dark energy equation of state {ω ν , $w_a$}. In this paper, we present the final emulator for the matter power spectrum based on 111 cosmological simulations, each covering a (2.1 Gpc) 3 volume and evolving 3200 3 particles. In this work, an additional set of 1776 lower resolution simulations and TimeRG perturbation theory results for the power spectrum are used to cover scales straddling the linear to mildly non-linear regimes (maximum wavenumber $\textit{k}$ = 5 Mpc –1 ). The emulator provides predictions at the 2–3 percent level of accuracy over a wide range of cosmological parameters and is publicly released as part of this paper.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Influence of Local Water Vapor Analysis Uncertainty on Ensemble Forecasts of Tropical Cyclogenesis Using Hurricane Irma (2017) as a Testbed

Abstract Tropical cyclone formation is known to require abundant water vapor in the lower to middle troposphere within the incipient disturbance. In this study, we assess the impacts of local water vapor analysis uncertainty on the predictability of the formation of Hurricane Irma (2017). To this end, we reduce the magnitude of the incipient disturbance’s water vapor perturbations obtained from an ensemble-based data assimilation system that constrained moisture by assimilating all-sky infrared and microwave radiances. Five-day ensemble forecasts are initialized two days before genesis using each set of modified analysis perturbations. Growth of convective differences and intensity uncertainty are evaluated for each ensemble forecast. We observe that when initializing an ensemble forecast with only moisture uncertainty within the incipient disturbance, the resulting intensity uncertainty at every lead time exceeds half that of an ensemble containing initial perturbations to all variables throughout the domain. Although ensembles with different initial moisture uncertainty amplitudes reveal a similar pathway to genesis, uncertainty in genesis timing varies substantially across ensembles since moister members exhibit earlier spinup of the low-level vortex. These differences in genesis timing are traced back to the first 6–12 h of integration, when differences in the position and intensity of mesoscale convective systems across ensemble members develop more quickly with greater initial moisture uncertainty. In addition, the rapid growth of intensity uncertainty may be greatly modulated by the diurnal cycle. Ultimately, this study underscores the importance of targeting the incipient disturbance with high spatiotemporal water vapor observations for ingestion into data assimilation systems. Significance Statement Hurricanes form from clusters of thunderstorms that organize into a coherent system. One of the key ingredients for the formation process is an abundance of moisture. In this study, we test the sensitivity of hurricane formation to the initial moisture content in the vicinity of the cluster of thunderstorms that would become Hurricane Irma (2017). To do so, we initialize sets of forecasts each having a different variability of initial moisture content within the embryonic disturbance. Our results show that the predictability of hurricane formation is highly dependent on the uncertainty of the moisture content within the initial disturbance. Consequently, more high-quality observations of the moisture within the precursor disturbances to hurricanes are expected to improve forecasts of their formation.

Hartman, Christopher M.↗
DOE OSTI2022

National and subnational short-term forecasting of COVID-19 in Germany and Poland during early 2021

During the COVID-19 pandemic there has been a strong interest in forecasts of the short-term development of epidemiological indicators to inform decision makers. In this study we evaluate probabilistic real-time predictions of confirmed cases and deaths from COVID-19 in Germany and Poland for the period from January through April 2021. We evaluate probabilistic real-time predictions of confirmed cases and deaths from COVID-19 in Germany and Poland. These were issued by 15 different forecasting models, run by independent research teams. Moreover, we study the performance of combined ensemble forecasts. Evaluation of probabilistic forecasts is based on proper scoring rules, along with interval coverage proportions to assess calibration. The presented work is part of a pre-registered evaluation study. We find that many, though not all, models outperform a simple baseline model up to four weeks ahead for the considered targets. Ensemble methods show very good relative performance. The addressed time period is characterized by rather stable non-pharmaceutical interventions in both countries, making short-term predictions more straightforward than in previous periods. However, major trend changes in reported cases, like the rebound in cases due to the rise of the B.1.1.7 (Alpha) variant in March 2021, prove challenging to predict. Multi-model approaches can help to improve the performance of epidemiological forecasts. However, while death numbers can be predicted with some success based on current case and hospitalization data, predictability of case numbers remains low beyond quite short time horizons. Additional data sources including sequencing and mobility data, which were not extensively used in the present study, may help to improve performance.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Methods for Incorporating Model Uncertainty into Exoplanet Atmospheric Analysis

A key goal of exoplanet spectroscopy is to measure atmospheric properties, such as abundances of chemical species, in order to connect them to our understanding of atmospheric physics and planet formation. In this new era of high-quality JWST data, it is paramount that these measurement methods are robust. When comparing atmospheric models to observations, multiple candidate models may produce reasonable fits to the data. Typically, conclusions are reached by selecting the best-performing model according to some metric. This ignores model uncertainty in favor of specific model assumptions, potentially leading to measured atmospheric properties that are overconfident and/or incorrect. In this paper, we compare three ensemble methods for addressing model uncertainty by combining posterior distributions from multiple analyses: Bayesian model averaging, a variant of Bayesian model averaging using leave-one-out predictive densities, and stacking of predictive distributions. We demonstrate these methods by fitting the Hubble Space Telescope (HST) + Spitzer transmission spectrum of the hot Jupiter HD 209458b using models with different cloud and haze prescriptions. All of our ensemble methods lead to uncertainties on retrieved parameters that are larger but more realistic and consistent with physical and chemical expectations. Since they have not typically accounted for model uncertainty, uncertainties of retrieved parameters from HST spectra have likely been underreported. We recommend stacking as the most robust model combination method. Our methods can be used to combine results from independent retrieval codes and from different models within one code. They are also widely applicable to other exoplanet analysis processes, such as combining results from different data reductions.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Data Imbalance, Uncertainty Quantification, and Transfer Learning in Data‐Driven Parameterizations: Lessons From the Emulation of Gravity Wave Momentum Transport in WACCM

Abstract Neural networks (NNs) are increasingly used for data‐driven subgrid‐scale parameterizations in weather and climate models. While NNs are powerful tools for learning complex non‐linear relationships from data, there are several challenges in using them for parameterizations. Three of these challenges are (a) data imbalance related to learning rare, often large‐amplitude, samples; (b) uncertainty quantification (UQ) of the predictions to provide an accuracy indicator; and (c) generalization to other climates, for example, those with different radiative forcings. Here, we examine the performance of methods for addressing these challenges using NN‐based emulators of the Whole Atmosphere Community Climate Model (WACCM) physics‐based gravity wave (GW) parameterizations as a test case. WACCM has complex, state‐of‐the‐art parameterizations for orography‐, convection‐, and front‐driven GWs. Convection‐ and orography‐driven GWs have significant data imbalance due to the absence of convection or orography in most grid points. We address data imbalance using resampling and/or weighted loss functions, enabling the successful emulation of parameterizations for all three sources. We demonstrate that three UQ methods (Bayesian NNs, variational auto‐encoders, and dropouts) provide ensemble spreads that correspond to accuracy during testing, offering criteria for identifying when an NN gives inaccurate predictions. Finally, we show that the accuracy of these NNs decreases for a warmer climate (4 × CO 2 ). However, their performance is significantly improved by applying transfer learning, for example, re‐training only one layer using ∼1% new data from the warmer climate. The findings of this study offer insights for developing reliable and generalizable data‐driven parameterizations for various processes, including (but not limited to) GWs.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Theory of Two-Level Tunneling Systems in Superconductors

We develop a field theory formulation for the interaction of an ensemble of two-level tunneling systems (TLSs) with the electronic states of a superconductor. Predictions for the impact of two-level tunneling systems on superconductivity are presented, including T c and the spectrum of quasiparticle states for conventional BCS superconductors. We show that nonmagnetic TLS impurities in conventional s-wave superconductors can act as pair-breaking or pair-enhancing defects depending on the level population of the distribution of TLS impurities. We present calculations of the enhancement of superconductivity, both T c and the order parameter, for TLS defects in thermal equilibrium with the electrons and lattice. The scattering of quasiparticles by TLS impurities leads to subgap states below the bulk excitation gap, Δ, as well as resonances in the continuum above Δ. The energies and spectral weights of these states depend on the distribution of tunnel splittings, while the spectral weights are particularly sensitive to the level occupation of the TLS impurities. Under microwave excitation, or decoupling from the thermal bath, a nonequilibrium level population of the TLS distribution generates subgap quasiparticle states near the Fermi level that contribute to dissipation and thus degrade the performance of superconducting devices at low temperatures.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2024

Modeling stochastic fluctuations in relativistic kinetic theory

Using the information current, we develop a Lorentz-covariant framework for modeling equilibrium fluctuations in relativistic kinetic theory in the grand-canonical ensemble. The resulting stochastic theory is proven to be causal and covariantly stable, and its predictions do not depend on the choice of spacetime foliation used to define the grand-canonical probabilities. As expected, in a box containing N > 5 particles, Boltzmann’s molecular chaos postulate is broken with (almost exact) probability N -1/2 , leading to a breakdown of the Boltzmann equation in small systems. Here, we also verify that, in ultrarelativistic gases, transient hydrodynamics already accounts for at least 80% of the equilibrium fluctuations of the stress-energy tensor at a given time. Finally, we compute the correlators at nonequal times for two selected collision kernels: that of a chemically active diluted solution, and that of ultrarelativistic scalar particles self-interacting via a quartic potential. For the former, we compute the density-density correlators analytically in real space, and dehydrodynamization of the stochastic theory is proven to occur whenever the mean free path diverges at high energy.

Astronomy & Astrophysics↗
DOE OSTI2025

Quantifying uncertainty in machine learning for nuclear binding energy

Techniques from artificial intelligence and machine learning are increasingly employed in nuclear theory; however, the uncertainties that arise from the complex parameter manifold encoded by the neural networks are often overlooked. Epistemic uncertainties arising from training the same network multiple times for an ensemble of initial weight sets offer a first insight into the confidence of machine learning predictions, but they often come with a high computational cost. Instead, we apply a single-model uncertainty quantification method called Δ-UQ that gives epistemic uncertainties with one-time training. Here, we demonstrate our approach on a two-feature model of nuclear binding energies per nucleon with proton and neutron number pairs as inputs. We show that Δ-UQ can produce reliable and self-consistent epistemic uncertainty estimates and can be used to assess the degree of confidence in predictions made with deep neural networks.

Huang, Mengyao [Lawrence Livermore National Labora↗
DOE OSTI2024

FORESTR: Finding, Organizing, Representing, Explaining, Summarizing, and Thinning Random forests

Random forests have become popular models used for data driven predictions. As a result, random forests are currently used or being considered for high-consequence mission applications in national security, such as the prediction of yield from optical signals and malware detection. While random forests may provide accurate predictions, the complexity of the algorithm causes a lack of interpretability. Random forests are an ensemble of regression or decision trees. Individual regression and decision trees are interpretable, but ensembles are inherently difficult to interpret due to the compilation of many models. We aim to increase the interpretability of random forests by finding patterns in the ensemble of trees that can be used to “thin” (or remove) trees. As a starting point, in this report, we develop a new distance metric for quantifying the similarity between trees based on their topologies (i.e., shapes). We base the metric on a novel distance metric for graphs that is a proper mathematical distance, is invariant to transformations, has registration between graphs, and computes topological evolutions between graphs. We use the tree distance metric to compute tree statistics such as a “mean tree” and to identify clusters of trees. We apply the developed methodology to a toy dataset and a mission relevant product inspection dataset to demonstrate how the metric can provide insight into random forests. Furthermore, we discuss the limitations of the approach and ideas for future research into how the metric could be used as a thinning tool to develop less complex models.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

A computational pipeline to generate a synthetic dataset of metal ion sorption to oxides for AI/ML exploration

The charged mineral/electrolyte interfaces are ubiquitous in the surface and subsurface–including the surroundings of the geological disposal sites for radioactive waste. Therefore, understanding how ions interact with charged surfaces is critically important for predicting radionuclide mobility in the case of waste leakage. At present, the Surface Complexation Models (SCMs) are the most successful thermodynamic frameworks to describe ion retention by mineral surfaces. SCMs are interfacial speciation models that account for the effect of the electric field generated by charged surfaces on sorption equilibria. These models have been successfully used to analyze and interpret a broad range of experimental observations including potentiometric and electrokinetic titrations or spectroscopy. Unfortunately, many of the current procedures to solve and fit SCM to experimental data are not optimal, which leads to a non-transferable or non-unique description of interfacial electrostatics and consequently of the strength and extent of ion retention by mineral surfaces. Recent developments in Artificial Intelligence (AI) offer a new avenue to replace SCM solvers and fitting algorithms with trained AI surrogates. Unfortunately, there is a lack of a standardized dataset covering a wide range of SCM parameter values available for AI exploration and training–a gap filled by this study. Here, we described the computational pipeline to generate synthetic SCM data and discussed approaches to transform this dataset into AI-learnable input. First, we used this pipeline to generate a synthetic dataset of electrostatic properties for a broad range of the prototypical oxide/electrolyte interfaces. The next step is to extend this dataset to include complex radionuclide sorption and complexation, and finally, to provide trained AI architectures able to infer SCMs parameter values rapidly from experimental data. Here, we illustrated the AI-surrogate development using the ensemble learning algorithms, such as Random Forest and Gradient Boosting. These surrogate models allow a rapid prediction of the SCM model parameters, do not rely on an initial guess, and guarantee convergence in all cases.

Li, Chunhui↗
DOE OSTI2022

Quantifying Northern High Latitude Gross Primary Productivity (GPP) Using Carbonyl Sulfide (OCS)

The northern high latitude (NHL, 40°N to 90°N) is where the second peak region of gross primary productivity (GPP) other than the tropics. The summer NHL GPP is about 80% of the tropical peak, but both regions are still highly uncertain. Carbonyl sulfide (OCS) provides an important proxy for photosynthetic carbon uptake. Here we optimize the OCS plant uptake fluxes across the NHL by fitting atmospheric concentration simulation with the GEOS-CHEM global transport model to the aircraft profiles acquired over Alaska during NASA's Carbon in Arctic Reservoirs Vulnerability Experiment (2012–2015). We use the empirical biome-specific linear relationship between OCS plant uptake flux and GPP to derive the six plant uptake OCS fluxes from different GPP data. Such GPP-based fluxes are used to drive the concentration simulations. We evaluate the simulations against the independent observations at two ground sites of Alaska. The optimized OCS fluxes suggest the NHL plant uptake OCS flux of -247 Gg S year -1 , about 25% stronger than the ensemble mean of the six GPP-based OCS fluxes. GPP-based OCS fluxes systematically underestimate the peak growing season across the NHL, while a subset of models predict early start of season in Alaska, consistent with previous studies of net ecosystem exchange. The OCS optimized GPP of 34 PgC yr -1 for NHL is also about 25% more than the ensembles mean from six GPP data. Further work is needed to fully understand the environmental and biotic drivers and quantify their rate of photosynthetic carbon uptake in Arctic ecosystems.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Dimensionally Reduced Model for Rapid and Accurate Prediction of Gas Saturation, Pressure, and Brine Production in a CO 2 Storage Application: Case Study Using the SACROC Field as Part of SMART Task 5

This technical report presents work conducted by the sub surface analysis team of the Strategic Systems Analysis & Engineering group at NETL for Task 5 of SMART Phase 1. This study involved the development of deep learning models for CO 2 geologic storage that are capable of accurate prediction of spatio-temporal outputs of CO 2 saturation, pressure, and brine production in three dimensional space over a storage operation's injection and post-injection timeframes. The model framework involves ensembling multi-layer encoder networks that provide dimesionality reduction of geologic inputs with fully connected long short-term memory (LSTM) neural networks that generate time-series prediction This approach offers a means to maximize training time efficiency, reduce computational memory burden, and minimize prediction turnaround.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Ab Initio Modeling on The Thermodynamic and Temperature-Dependent Elastic Properties of Subsystems of The FCC FeNiCoCr Medium Entropy Alloys (MEAs)

The stability, phonon spectra, thermodynamic, and temperature-dependent elasticity of subsystems of the FCC FeNiCoCr MEAs are systematically studied by the ab initio approach. Especially, the quasi-harmonic approximation (QHA) and the innovative Zentropy theory were utilized to predict the thermodynamic properties and elastic properties of FeNi, NiCo, FeNiCo, and FeNiCoCr MEAs with the consideration of magnetic transition. With the ensemble of the partition function based on the multiplicity of each magnetic configuration, the current work successfully reproduced the Curie temperature and the Schottky anomaly of heat capacity of these four MEAs purely based on the ab initio input. Meanwhile, the elastic properties of these alloys at finite temperatures are also successfully predicted with the consideration of magnetic transition. The overall results agree well with the available experimental data and CALPHAD prediction.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Evaluating the performance of random forest and iterative random forest based methods when applied to gene expression data

Gene-to-gene networks, such as Gene Regulatory Networks (GRN) and Predictive Expression Networks (PEN) capture relationships between genes and are beneficial for use in downstream biological analyses. There exists multiple network inference tools to produce these gene-to-gene networks from matrices of gene expression data. Random Forest-Leave One Out Prediction (RF-LOOP) is a method that has been shown to be efficient at producing these gene-to-gene networks, frequently known as GEne Network Inference with Ensemble of trees (GENIE3). Random Forest can be replaced in this process by iterative Random Forest (iRF), which performs variable selection and boosting. Here we validate that iterative Random Forest-Leave One Out Prediction (iRF-LOOP) produces higher quality networks than GENIE3 (RF-LOOP). We use both synthetic and empirical networks from the Dialogue for Reverse Engineering Assessment and Methods (DREAM) Challenges by Sage Bionetworks, as well as two additional empirical networks created from Arabidopsis thaliana and Populus trichocarpa expression data.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Scaling Ensembles of Data-Intensive Quantum Chemical Calculations for Millions of Molecules

Deep learning models are efficient computational tools that can accelerate the inverse design of molecules with desired functional properties by generating predictions at a fraction of the time required by traditional quantum chemical approaches. To ensure that a model maintains accuracy and transferability across broad regions of the chemical space explored during the inverse design, it must be trained on massively large volumes of simulation data. This requires running large-scale ensemble quantum chemical calculations on high-performance computing (HPC) systems for data collection. However, the efficient execution of such large ensemble calculations and the management of large volumes of output data require tools that can judiciously utilize computational resources and manage metadata overhead on the file system. Therefore, we present a high-performance, scalable, ensemble management framework for performing data-intensive quantum chemical electronic structure calculations for organic molecules. This framework provides abstractions to plug different ab initio, first principles, and first principles-based semi-empirical methods and executes them efficiently at large scale on HPC systems. It dynamically distributes tasks to resources and uses tiered storage for managing large collections of files. We employed this framework to process over ten million organic molecules and generate open-source datasets that provide UV-vis absorption spectra by running time-dependent density-functional tight-binding calculations. It is the largest database containing molecular optical spectra that were simulated with quantum chemical methods in a consistent manner.

Mehta, Kshitij↗
DOE OSTI2021

Scalable statistical inference of photometric redshift via data subsampling

Handling big data has largely been a major bottleneck in traditional statistical models. Consequently, when accurate point prediction is the primary target, machine learning models are often preferred over their statistical counterparts for bigger problems. But full probabilistic statistical models often outperform other models in quantifying uncertainties associated with model predictions. We develop a data-driven statistical modeling framework that combines the uncertainties from an ensemble of statistical models learned on smaller subsets of data carefully chosen to account for imbalances in the input space. We demonstrate this method on a photometric redshift estimation problem in cosmology, which seeks to infer a distribution of the redshift—the stretching effect in observing the light of far-away galaxies—given multivariate color information observed for an object in the sky. Our proposed method performs balanced partitioning, graph-based data subsampling across the partitions, and training of an ensemble of Gaussian process models.

data subsampling↗