Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “error estimation”

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

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

At least 1,045 records · Page 58

Constraining Dark Matter in Galactic Substructure

Detecting the dark matter annihilation signal from Galactic substructure, or subhalos, is an important challenge for high-energy gamma-ray experiments. In this paper we discuss detection prospects by combining two different aspects of the gamma-ray signal: the angular distribution and the photon counts probability distribution function (PDF). The true PDF from subhalos has been shown recently (by Lee et al.) to deviate from Poisson/ we extend this analysis and derive the signal PDF from a detailed Lambda-CDM-based model for the properties of subhalos. We combine our PDF with a model for Galactic and extra-Galactic diffuse gamma-ray emission to obtain an estimator and projected error on dark matter particle properties (mass and annihilation cross section) using the Fermi Gamma-Ray Space Telescope. We compare the estimator obtained from the true PDF to that obtained from the simpler Poisson analysis. We find that, although both estimators are unbaised in the presence of backgrounds, the error on dark matter properties derived from the true PDF is ~50% smaller than when utilizing the Poisson-based analysis.

79 ASTRONOMY AND ASTROPHYSICS↗

An accurate symplectic calculation of the inboard magnetic footprint from statistical topological noise and field errors in the DIII-D

Any canonical transformation of Hamiltonian equations is symplectic, and any area-preserving transformation in 2D is a symplectomorphism. Based on these, a discrete symplectic map and its continuous symplectic analog are derived for forward magnetic field line trajectories in natural canonical coordinates. The unperturbed axisymmetric Hamiltonian for magnetic field lines is constructed from the experimental data in the DIII-D [J. L. Luxon and L. E. Davis, Fusion Technol. 8, 441 (1985)]. The equilibrium Hamiltonian is a highly accurate, analytic, and realistic representation of the magnetic geometry of the DIII-D. These symplectic mathematical maps are used to calculate the magnetic footprint on the inboard collector plate in the DIII-D. Internal statistical topological noise and field errors are irreducible and ubiquitous in magnetic confinement schemes for fusion. It is important to know the stochasticity and magnetic footprint from noise and error fields. The estimates of the spectrum and mode amplitudes of the spatial topological noise and magnetic errors in the DIII-D are used as magnetic perturbation. The discrete and continuous symplectic maps are used to calculate the magnetic footprint on the inboard collector plate of the DIII-D by inverting the natural coordinates to physical coordinates. The combination of highly accurate equilibrium generating function, natural canonical coordinates, symplecticity, and small step-size together gives a very accurate calculation of magnetic footprint. Radial variation of magnetic perturbation and the response of plasma to perturbation are not included. The inboard footprint from noise and errors are dominated by m=3, n=1 mode. The footprint is in the form of a toroidally winding helical strip. The width of stochastic layer scales as (1/2) power of amplitude. The area of footprint scales as first power of amplitude. The physical parameters such as toroidal angle, length, and poloidal angle covered before striking, and the safety factor all have fractal structure. The average field diffusion near the X-point for lines that strike and that do not strike differs by about three to four orders of magnitude. The magnetic footprint gives the maximal bounds on size and heat flux density on collector plate.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Assessment of Subyearling Chinook Salmon Survival through the Federal Hydropower Projects in the Main-Stem Columbia River

High survival through hydropower projects is an essential element in the recovery of salmonid populations in the Columbia River. It is also a regulatory requirement under the 2008 Federal Columbia River Power System (FCRPS) Biological Opinion (BiOp) established under the Endangered Species Act. It requires dam passage survival to be ≥0.96 and ≥0.93 for spring and summer outmigrating juvenile salmonids, respectively, and estimated with a standard error ≤ 0.015. An innovative virtual/paired-release design was used to estimate dam passage survival, defined as survival from the face of a dam to the tailrace mixing zone. A coordinated four-dam study was conducted during the 2012 summer outmigration using 14,026 run-of-river subyearling Chinook salmon surgically implanted with acoustic micro-transmitter (AMT) tags released at 9 different locations, and monitored on 14 different detection arrays. Each of the four estimates of dam passage survival exceeded BiOp requirements with values ranging from 0.9414 to 0.9747 and standard errors, 0.0031 to 0.0114. Two consecutive years of survival estimates must meet BiOp standards in order for a hydropower project to be in compliance with recovery requirements for a fish stock.

Skalski, J. R.↗

Collimator design for a dedicated molecular breast imaging-guided biopsy system: Proof-of-concept

Purpose: Molecular breast imaging (MBI) is a dedicated nuclear medicine breast imaging modality that employs dual-head cadmium zinc telluride (CZT) gamma cameras to functionally detect breast cancer. MBI has been shown to detect breast cancers otherwise occult on mammography and ultrasound. Currently, a MBI-guided biopsy system does not exist to biopsy such lesions. Our objective was to consider the utility of a novel conical slant-hole (CSH) collimator for rapid (<1 min) and accurate monitoring of lesion position to serve as part of a MBI-guided biopsy system. Methods: An initial CSH collimator design was derived from the dimensions of a parallel-hole collimator optimized for MBI performed with dual-head CZT gamma cameras. The parameters of the CSH collimator included the collimator height, cone slant angle, thickness of septa and cones of the collimator, and the annular areas exposed at the base of the cones. These parameters were varied within the geometric constraints of the MBI system to create several potential CSH collimator designs. The CSH collimator designs were evaluated using Monte Carlo simulations. The model included a breast compressed to a thickness of 6 cm with a 1-cm diameter lesion located 3 cm from the collimator face. The number of particles simulated was chosen to represent the count density of a low-dose, screening MBI study acquired with the parallel-hole collimator for 10 min after a {approx}150 MBq (4 mCi) injection of Tc-99m sestamibi. The same number of particles was used for the CSH collimator simulations. In the resulting simulated images, the count sensitivity, spatial resolution, and accuracy of the lesion depth determined from the lesion profile width were evaluated. Results: The CSH collimator design with default parameters derived from the optimal parallel-hole collimator provided 1-min images with error in the lesion depth estimation of 1.1 {+-} 0.7 mm and over 21 times the lesion count sensitivity relative to 1-min images acquired with the current parallel-hole collimator. Sensitivity was increased via more vertical cone slant angles, larger annular areas, thinner cone walls, shorter cone heights, and thinner radiating septa. Full width at half maximum trended in the opposite direction as sensitivity for all parameters. There was less error in the depth estimates for less vertical slant angles, smaller annular areas, thinner cone walls, cone heights near 1 cm, and generally thinner radiating septa. Conclusions: A Monte Carlo model was used to demonstrate the feasibility of a CSH collimator design for rapid biopsy application in molecular breast imaging. Specifically, lesion depth of a 1-cm diameter lesion positioned in the center of a typical breast can be estimated with error of less than 2 mm using circumferential count profiles of images acquired in 1 min.

60 APPLIED LIFE SCIENCES↗

Towards optimal sensor placement for inverse problems in spaces of measures

The objective of this work is to quantify the reconstruction error in sparse inverse problems with measures and stochastic noise, motivated by optimal sensor placement. To be useful in this context, the error quantities must be explicit in the sensor configuration and robust with respect to the source, yet relatively easy to compute in practice, compared to a direct evaluation of the error by a large number of samples. In particular, we consider the identification of a measure consisting of an unknown linear combination of point sources from a finite number of measurements contaminated by Gaussian noise. The statistical framework for recovery relies on two main ingredients: first, a convex but non-smooth variational Tikhonov point estimator over the space of Radon measures and, second, a suitable mean-squared error based on its Hellinger–Kantorovich distance to the ground truth. To quantify the error, we employ a non-degenerate source condition as well as careful linearization arguments to derive a computable upper bound. This leads to asymptotically sharp error estimates in expectation that are explicit in the sensor configuration. Thus they can be used to estimate the expected reconstruction error for a given sensor configuration and guide the placement of sensors in sparse inverse problems.

97 MATHEMATICS AND COMPUTING↗

Evaluation of multiple image-based modalities for image-guided radiation therapy (IGRT) of prostate carcinoma: A prospective study

Purpose: Setup errors and prostate intrafraction motion are main sources of localization uncertainty in prostate cancer radiation therapy. This study evaluates four different imaging modalities 3D ultrasound (US), kV planar images, cone-beam computed tomography (CBCT), and implanted electromagnetic transponders (Calypso/Varian) to assess inter- and intrafraction localization errors during intensity-modulated radiation therapy based treatment of prostate cancer. Methods: Twenty-seven prostate cancer patients were enrolled in a prospective IRB-approved study and treated to a total dose of 75.6 Gy (1.8 Gy/fraction). Overall, 1100 fractions were evaluated. For each fraction, treatment targets were localized using US, kV planar images, and CBCT in a sequence defined to determine setup offsets relative to the patient skin tattoos, intermodality differences, and residual errors for each patient and patient cohort. Planning margins, following van Herk's formalism, were estimated based on error distributions. Calypso-based localization was not available for the first eight patients, therefore centroid positions of implanted gold-seed markers imaged prior to and immediately following treatment were used as a motion surrogate during treatment. For the remaining 19 patients, Calypso transponders were used to assess prostate intrafraction motion. Results: The means ({mu}), and standard deviations (SD) of the systematic ({Sigma}) and random errors ({sigma}) of interfraction prostate shifts (relative to initial skin tattoo positioning), as evaluated using CBCT, kV, and US, averaged over all patients and fractions, were: [{mu}{sub CBCT}= (-1.2, 0.2, 1.1) mm, {Sigma}{sub CBCT}= (3.0, 1.4, 2.4) mm, {sigma}{sub CBCT}= (3.2, 2.2, 2.5) mm], [{mu}{sub kV}= (-2.9, -0.4, 0.5) mm, {Sigma}{sub kV}= (3.4, 3.1, 2.6) mm, {sigma}{sub kV}= (2.9, 2.0, 2.4) mm], and [{mu}{sub US}= (-3.6, -1.4, 0.0) mm, {Sigma}{sub US}= (3.3, 3.5, 2.8) mm, {sigma}{sub US}= (4.1, 3.8, 3.6) mm], in the anterior-posterior (A/P), superior-inferior (S/I), and the left-right (L/R) directions, respectively. In the treatment protocol, adjustment of couch was guided by US images. Residual setup errors as assessed by kV images were found to be: {mu}{sub residual}= (-0.4, 0.2, 0.2) mm, {Sigma}{sub residual}= (1.0, 1.0,0.7) mm, and {sigma}{sub residual}= (2.5, 2.3, 1.8) mm. Intrafraction prostate motion, evaluated using electromagnetic transponders, was: {mu}{sub intrafxn}= (0.0, 0.0, 0.0) mm, {Sigma}{sub intrafxn}= (1.3, 1.5, 0.6) mm, and {sigma}{sub intrafxn}= (2.6, 2.4, 1.4) mm. Shifts between pre- and post-treatment kV images were: {mu}{sub kV(post-pre)}= (-0.3, 0.8, -0.2), {Sigma}{sub kV(post-pre)}= (2.4, 2.7, 2.1) mm, and {sigma}{sub kV(post-pre)}= (2.7, 3.2, 3.1) mm. Relative to skin tattoos, planning margins for setup error were within 10-11 mm for all image-based modalities. The use of image guidance was shown to reduce these margins to less than 5 mm. Margins to compensate for both residual setup (interfraction) errors as well as intrafraction motion were 6.6, 6.8, and 3.9 mm in the A/P, S/I, and L/R directions, respectively. Conclusions: Analysis of interfraction setup errors, performed with US, CBCT, planar kV images, and electromagnetic transponders, from a large dataset revealed intermodality shifts were comparable (within 3-4 mm). Interfraction planning margins, relative to setup based on skin marks, were generally within the 10 mm prostate-to-planning target volume margin used in our clinic. With image guidance, interfraction residual planning margins were reduced to approximately less than 4 mm. These findings are potentially important for dose escalation studies using smaller margins to better protect normal tissues.

60 APPLIED LIFE SCIENCES↗

SU-F-T-29: The Important of Each Fraction Image-Guided Planning for Postoperative HDR-Brachytherapy in Endometrial Carcinoma

Purpose: Cylindrical applicators are often used for postoperative HDRbrachytherapy in endometrial carcinoma. It has been considered that dosimetric variation between fractions for this treatment is minimal and might not be necessary to perform treatment planning for every fractions. At our institute, it is traditional to perform treatment planning with CT simulation on the first fraction and uses this plan for the rest of treatment. This study was aim to evaluate the errors of critical structure doses between the fractions when simulation and planning were done for first fraction only. Methods: Treatment plans of 10 endometrial carcinoma patients who received postoperative HDR-brachytherapy and underwent CT-simulation for every HDR-fractions at our department were retrospectively reviewed. All of these patients were treated with cylindrical applicator and prescribed dose 15Gy in 3 fractions to 0.5cm from vaginal surface. The treatment plan from the first fraction was used to simulate in second and third CT-simulation. Radiation dose for critical structures in term of Dose-to-2cc (D2cc) were evaluated and compared between planning CT. Results: The D2cc for bladder and rectum were evaluated. For bladder, the mean error of D2cc estimation for second and third fractions was 7.6% (0.1–20.1%, SD=5.7). And the mean error for D2cc of rectum was 8.5% (0.1–29.4%, SD=8.5). Conclusion: The critical structure doses could be significant difference between fractions which may affects treatment outcomes or toxicities. From our data, image-guided brachytherapy at least with CT-Simulation should be done for every treatment fractions.

60 APPLIED LIFE SCIENCES↗

Determining mutant spectra of three RNA viral samples using ultra-deep sequencing

RNA viruses have extremely high mutation rates that enable the virus to adapt to new host environments and even jump from one species to another. As part of a viral transmission study, three viral samples collected from naturally infected animals were sequenced using Illumina paired-end technology at ultra-deep coverage. In order to determine the mutant spectra within the viral quasispecies, it is critical to understand the sequencing error rates and control for false positive calls of viral variants (point mutantations). I will estimate the sequencing error rate from two control sequences and characterize the mutant spectra in the natural samples with this error rate.

59 BASIC BIOLOGICAL SCIENCES↗

INFERRING THE GALACTIC POTENTIAL WITH GAIA AND FRIENDS: SYNERGIES WITH OTHER SURVEYS

In the coming decade, the Gaia satellite will precisely measure the positions and velocities of millions of stars in the Galactic halo, including stars in many tidal streams. These streams, the products of hierarchical accretion of satellite galaxies by the Milky Way (MW), can be used to infer the Galactic gravitational potential thanks to their initial compactness in phase space. Plans for observations to extend Gaia’s radial velocity (RV) measurements to faint stars, and to determine precise distances to RR Lyrae in streams, would further extend the power of Gaia’s kinematic catalog to characterize the MW’s potential at large Galactocentric distances. In this work I explore the impact of these extra data on the ability to fit the potential using the method of action clustering, which statistically maximizes the information content (clumpiness) of the action space of tidal streams, eliminating the need to determine stream membership for individual stars. Using a mock halo in a toy spherical potential, updated post-launch error models for Gaia, and estimates for RV and distance errors for the tracers to be followed up, I show that combining either form of additional information with the Gaia catalog greatly reduces the bias in determining the scale radius and total mass of the Galaxy compared to the use of Gaia data alone.

79 ASTRONOMY AND ASTROPHYSICS↗

A continuation multilevel Monte Carlo algorithm

Here, we propose a novel Continuation Multi Level Monte Carlo (CMLMC) algorithm for weak approximation of stochastic models. The CMLMC algorithm solves the given approximation problem for a sequence of decreasing tolerances, ending when the required error tolerance is satisfied. CMLMC assumes discretization hierarchies that are defined a priori for each level and are geometrically refined across levels. Moreover, the actual choice of computational work across levels is based on parametric models for the average cost per sample and the corresponding variance and weak error. These parameters are calibrated using Bayesian estimation, taking particular notice of the deepest levels of the discretization hierarchy, where only few realizations are available to produce the estimates. The resulting CMLMC estimator exhibits a non-trivial splitting between bias and statistical contributions. We also show the asymptotic normality of the statistical error in the MLMC estimator and justify in this way our error estimate that allows prescribing both required accuracy and confidence in the final result. Our numerical results substantiate the above results and illustrate the corresponding computational savings in examples that are described in terms of differential equations either driven by random measures or with random coefficients.

97 MATHEMATICS AND COMPUTING↗

Notes on Regression Analysis for Radar Parameter Estimation

A fundamental task of radar, beyond merely detecting a target, is to estimate some parameters associated with it. For example, this might include range, direction, velocity, etc. In any case, multiple measurements, often noisy, need to be processed to yield a ‘best estimate’ of the parameter. A common mathematical method for doing so is called “Regression” analysis. The goal is to minimize the expected squared error in the estimate. Even when alternate algorithms are considered, the least squared-error regression analysis is the benchmark against which alternatives are compared.

47 OTHER INSTRUMENTATION↗

Evaluation of GlassNet for physics-informed machine learning of glass stability and glass-forming ability

Glassy materials form the basis of many modern applications, including nuclear waste immobilization, touch-screen displays, and optical fibers, and also hold great potential for future medical and environmental applications. However, their structural complexity and large composition space make design and optimization challenging for certain applications. Of particular importance for glass processing and design is an estimate of a given composition's glass-forming ability (GFA). However, there remain many open questions regarding the underlying physical mechanisms of glass formation, especially in oxide glasses. It is apparent that a proxy for GFA would be highly useful in glass processing and design, but identifying such a surrogate property has proven itself to be difficult. While glass stability (GS) parameters have historically been used as a GFA surrogate, recent research has demonstrated that most of these parameters are not accurate predictors of the GFA of oxide glasses. Here, in this work, we explore the application of an open-source pre-trained neural network model, GlassNet, that can predict the characteristic temperatures necessary to compute GS with reasonable performance and assess the feasibility of using these physics-informed machine learning (PIML)-predicted GS parameters to estimate GFA. In doing so, we track the uncertainties at each step of the computation—from the original ML prediction errors to the compounding of errors during GS estimation, and finally to the final estimation of GFA. While GlassNet exhibits reasonable accuracy on all individual properties, we observe a large compounding of error in the combination of these individual predictions for the PIML prediction of GS, finding that random forest models offer similar accuracy to GlassNet. We also break down the performance of GlassNet on different glass families and find that the error in GS prediction is correlated with the error in crystallization peak temperature prediction. Lastly, we utilize this finding to assess the relationship between top-performing GS parameters and GFA for two ternary glass systems: sodium borosilicate and sodium iron phosphate glasses. We conclude that to obtain true ML predictive capability of GFA, significantly more data needs to be collected.

36 MATERIALS SCIENCE↗

Model correction and updating of a stochastic degradation model for failure prognostics of miter gates

Understanding the degradation of the quoin block is vital for failure prognostics in miter gates. Due to the complicated degradation mechanism, degradation models based on simplifications and assumptions cannot accurately describe the damage evolution. It is observed that small errors in a simplified degradation model can lead to a large discrepancy in the remaining useful life estimation attributed to error accumulation over time. Aiming to address this issue in failure prognostics, this paper presents a dynamic model correction framework for a simplified degradation model using strain measurements. In the proposed framework, a polynomial chaos expansion (PCE) model is employed to compensate the missing physics in a simplified stochastic degradation model. Here, a maximum likelihood estimation method is developed to estimate the uncertain parameters of the simplified physics-based degradation model along with the unknown PCE model parameters using strain measurements as the observables. The updated damage degradation model is then applied to failure prognostics of a miter gate. Results of a case study show that the proposed approach can effectively improve the accuracy of failure prognostics in miter gates.

42 ENGINEERING↗

Estimating the Bias of Local Polynomial Approximations Using the Peano Kernel

These presentation visuals define local polynomial approximations, give formulas for bias and random components of the error, and express bias error in terms of the Peano kernel. They further derive constants that give figures of merit, and show the figures of merit for 3 common weighting functions. The Peano kernel theorem yields estimates for the bias error for local-polynomial-approximation smoothing that are superior in several ways to the error estimates in the current literature.

97 MATHEMATICS AND COMPUTING↗

A Physics-Based Digital Twin for Wave Elevation and Seabed Moment Estimation of Offshore Monopiles: Preprint

In this work, we present a proof of concept of a physics-based digital twin for a monopile structure (with overhead inertia) subjected to wave loading. The digital twin is formulated using reduced-order models derived from first principles and combined with a Kalman filter for state estimation. The proposed framework estimates the monopile top motion, the wave elevation, and the section forces and moments along the pile using primarily acceleration measurements at the monopile top. Key innovations include the use of a hydrodynamic shape function to represent distributed wave loading in a compact and computationally efficient manner, and the introduction of a shaping filter to augment the state-space with wave kinematics. Synthetic measurement data are generated using OpenFAST and used as a reference to assess the performance of the digital twin. Results demonstrate that the wave elevation can be accurately reconstructed without direct sea-state measurements as long as the wave regime is inertia-dominated. Under the ideal tested conditions, the total hydrodynamic force and sea-bed bending moment are estimated with relative errors on the order of 1% and correlation coefficients exceeding 96%. Future work will evaluate the estimator's performance under operational uncertainties and more complex loading conditions.

17 WIND ENERGY↗

Estimating Aboveground Biomass in Tropical Forests: Field Methods and Error Analysis for the Calibration of Remote Sensing Observations

Mapping and monitoring of forest carbon stocks across large areas in the tropics will necessarily rely on remote sensing approaches, which in turn depend on field estimates of biomass for calibration and validation purposes. Here, we used field plot data collected in a tropical moist forest in the central Amazon to gain a better understanding of the uncertainty associated with plot-level biomass estimates obtained specifically for the calibration of remote sensing measurements. In addition to accounting for sources of error that would be normally expected in conventional biomass estimates (e.g., measurement and allometric errors), we examined two sources of uncertainty that are specific to the calibration process and should be taken into account in most remote sensing studies: the error resulting from spatial disagreement between field and remote sensing measurements (i.e., co-location error), and the error introduced when accounting for temporal differences in data acquisition. We found that the overall uncertainty in the field biomass was typically 25% for both secondary and primary forests, but ranged from 16 to 53%. Co-location and temporal errors accounted for a large fraction of the total variance (>65%) and were identified as important targets for reducing uncertainty in studies relating tropical forest biomass to remotely sensed data. Although measurement and allometric errors were relatively unimportant when considered alone, combined they accounted for roughly 30% of the total variance on average and should not be ignored. Lastly, our results suggest that a thorough understanding of the sources of error associated with field-measured plot-level biomass estimates in tropical forests is critical to determine confidence in remote sensing estimates of carbon stocks and fluxes, and to develop strategies for reducing the overall uncertainty of remote sensing approaches.

54 ENVIRONMENTAL SCIENCES↗

Convergence Study in Global Sensitivity Analysis

Monte Carlo (MC) sampling is a common method used to randomly sample a range of scenarios. The associated error follows a predictable rate of convergence of $1/\sqrt{N}$, such that quadrupling the sample size halves the error. This method is often employed in performing global sensitivity analysis which computes sensitivity indices, measuring fractional contributions of uncertain model inputs to the total output variance. In this study, several models are used to observe the rate of decay in the MC error in the estimation of the conditional variance, the total variance in the output, and the global sensitivity indices. The purpose is to examine the rate of convergence of the error in existing specialized, albeit MC-based, sampling methods for estimation of the sensitivity indices. It was found that the conditional variances and sensitivity indices all follow the $1/\sqrt{N}$ convergence rate. Future work will test the convergence of observables from more complex models such as ignition time in combustion.

97 MATHEMATICS AND COMPUTING↗

An adaptive Gaussian process-based method for efficient Bayesian experimental design in groundwater contaminant source identification problems: ADAPTIVE GAUSSIAN PROCESS-BASED INVERSION

Surrogate models are commonly used in Bayesian approaches such as Markov Chain Monte Carlo (MCMC) to avoid repetitive CPU-demanding model evaluations. However, the approximation error of a surrogate may lead to biased estimations of the posterior distribution. This bias can be corrected by constructing a very accurate surrogate or implementing MCMC in a two-stage manner. Since the two-stage MCMC requires extra original model evaluations, the computational cost is still high. If the information of measurement is incorporated, a locally accurate approximation of the original model can be adaptively constructed with low computational cost. Based on this idea, we propose a Gaussian process (GP) surrogate-based Bayesian experimental design and parameter estimation approach for groundwater contaminant source identification problems. A major advantage of the GP surrogate is that it provides a convenient estimation of the approximation error, which can be incorporated in the Bayesian formula to avoid over-confident estimation of the posterior distribution. The proposed approach is tested with a numerical case study. Without sacrificing the estimation accuracy, the new approach achieves about 200 times of speed-up compared to our previous work using two-stage MCMC.

Zhang, Jiangjiang↗