2,023 Hits in 3.5 sec

A comparative study of metamodeling methods considering sample quality merits

Dong Zhao, Deyi Xue
2010 Structural And Multidisciplinary Optimization  
In this work, four types of metamodeling methods, including multivariate polynomial method, radial basis function method, kriging method and Bayesian neural network method, three sample quality merits,  ...  In addition, the Bayesian neural network method, which is rarely used in metamodeling and has never been considered in comparative studies, is selected in this research as a metamodeling method and compared  ...  The impact of sample size is examined by using the Latin hypercube sampling method to generate uniformly scattered samples of different sizes in the design space.  ... 
doi:10.1007/s00158-010-0529-3 fatcat:4iwmqp4epvehdgbet7rut3gzoa

Efficient Exploration of Microstructure-Property Spaces via Active Learning

Lukas Morand, Norbert Link, Tarek Iraki, Johannes Dornheim, Dirk Helm
2022 Frontiers in Materials  
For virtual materials design, typically data is generated by numerical simulations, which implies that data pairs can be sampled on demand at arbitrary locations in microstructure space.  ...  In materials design, supervised learning plays an important role for optimization and inverse modeling of microstructure-property relations.  ...  on the Latin hypercube samples.  ... 
doi:10.3389/fmats.2021.824441 fatcat:sqauurn4ljekvapcui6tpp6l7q

A comparison of Latin hypercube and grid ensemble designs for the multivariate emulation of an Earth system model

Nathan M. Urban, Thomas E. Fricker
2010 Computers & Geosciences  
We also find that the Latin hypercube emulator is more accurate than the grid emulator in single-parameter model sensitivity studies.  ...  However, Latin hypercube designs have well known theoretical advantages in the design of computer experiments, especially as the dimension of the parameter space grows.  ...  Acknowledgments This study was inspired by work undertaken at the 2008 Probability and Uncertainty in Climate Modeling (PUCM) Research Playground workshop at Durham University.  ... 
doi:10.1016/j.cageo.2009.11.004 fatcat:ygrhqi74rfhbbnmpfhs3ny3ow4

Neural Network Emulation of Reionization Simulations

Claude J. Schmit, Jonathan R. Pritchard
2017 Proceedings of the International Astronomical Union  
We consider the use of an artificial neural network to emulate 21cmFAST simulations and use it in a Bayesian parameter inference study.  ...  We find that the use of a training set of size 100 samples can recover the error contours of a full scale MCMC analysis which evaluates the model at each step.  ...  Panel A shows the recovered contours for an ANN trained on 100 training sets sampled from a latin hypercube. Panel B shows the results when training the emulator on 1000 latin hypercube samples.  ... 
doi:10.1017/s174392131700984x fatcat:wimbbd5p2rdd3alw6t5yv3x2se

CONNECT: A neural network based framework for emulating cosmological observables and cosmological parameter inference [article]

Andreas Nygaard, Emil Brinch Holm, Steen Hannestad, Thomas Tram
2022 arXiv   pre-print
In this paper we present connect, a neural network framework emulating class computations as an easy-to-use plug-in for the popular sampler MontePython. connect uses an iteratively trained neural network  ...  Bayesian parameter inference is an essential tool in modern cosmology, and typically requires the calculation of 10^5–10^6 theoretical models for each inference of model parameters for a given dataset  ...  upon creation of the data when using Latin hypercube sampling (iterative sampling).  ... 
arXiv:2205.15726v1 fatcat:h2j47wqgovdghjzypcr5hpp6g4

Emulation of reionization simulations for Bayesian inference of astrophysics parameters using neural networks

C J Schmit, J R Pritchard
2017 Monthly notices of the Royal Astronomical Society  
We consider the use of artificial neural networks as a blind emulation technique.  ...  However, numerical modelling of the Epoch of Reionization can be prohibitively expensive for Bayesian parameter inference and how to optimally extract information from incoming data is currently unclear  ...  Latin Hypercube approach A second approch is to use the latin hypercube sampling (LHS) technique, shown in Figure 2 .  ... 
doi:10.1093/mnras/stx3292 fatcat:pdwc3ifc5jc5rg5pxuupw4zr4e

Learning to Warm-Start Bayesian Hyperparameter Optimization [article]

Jungtaek Kim, Saehoon Kim, Seungjin Choi
2018 arXiv   pre-print
To this end, we introduce a Siamese network composed of deep feature and meta-feature extractors, where deep feature extractor provides a semantic representation of each instance in a dataset and meta-feature  ...  Then, our learned meta-features are used to select a few datasets similar to the new dataset, so that hyperparameters in similar datasets are adopted as initializations to warm-start Bayesian hyperparameter  ...  Uniform, Latin, and Halton stand for näive uniform random sampling, Latin hypercube sampling, and quasi-Monte Carlo sampling with Halton sequence, respectively.  ... 
arXiv:1710.06219v3 fatcat:nqd3zduti5aehakgnfehbmqghe

Laminated Composites Buckling Analysis Using Lamination Parameters, Neural Networks and Support Vector Regression

Rubem M. Koide, Ana Paula C. S. Ferreira, Marco A. Luersen
2015 Latin American Journal of Solids and Structures  
Latin hypercube sampling in bayesian networks, In Proceedings of FLAIRS-2000- AAAI 2000, Orlando, USA. Erdal, O., Sonmez, F.O. (2005).  ...  As an example, a ten-point Latin hypercube sampling plan for a laminated of two layers is shown in Figure 2.  ... 
doi:10.1590/1679-78251237 fatcat:qk6huvzy5bajrntwgsl3iouogm

A knowledge-based system for numerical design of experiments processes in mechanical engineering

Gaëtan Blondet, Julien Le Duigou, Nassim Boudaoud
2019 Expert systems with applications  
. • A bayesian network with a "multi-net" strategy is proposed. • Models are trained from historical data and expert knowledge. • The performances of the proposed method are validated through a case study  ...  Acknowledgments This work is done in the French FUI project SDM4DOE. We also thank all consortium partners for their contribution during the development of ideas and concepts proposed in the paper.  ...  generation Latin Hypercube Type of DoE Sampling ; Halton sequence ; Sobol Type of sampling method to define each experiment.  ... 
doi:10.1016/j.eswa.2019.01.013 fatcat:7bybpjckobct7cofb6njvw33xa

Unbounded Bayesian Optimization via Regularization [article]

Bobak Shahriari and Alexandre Bouchard-Côté and Nando de Freitas
2015 arXiv   pre-print
In this work we modify the standard Bayesian optimization framework in a principled way to allow automatic resizing of the search space.  ...  and compare them on two common synthetic benchmarking test functions as well as the tasks of tuning the stochastic gradient descent optimizer of a multi-layered perceptron and a convolutional neural network  ...  We emphasize that in this work we have addressed one of the challenges that must be overcome toward the development of practical Bayesian optimization hyper-parameter tuning tools.  ... 
arXiv:1508.03666v1 fatcat:kxwrem25ufctpiywqr3esfitde

How do we optimally sample model grids of exoplanet spectra? [article]

Chloe Fisher, Kevin Heng
2022 arXiv   pre-print
Here we investigate alternative methods of sampling parameters, including random sampling and Latin hypercube (LH) sampling, and how these compare to linearly sampled grids.  ...  Our results show that random or LH sampling out-performs linear sampling in parameter predictability for our higher dimensional models, requiring fewer models in the grid, and thus allowing for more computationally  ...  Latin Hypercube Sampling A common sampling technique in machine learning is Latin hypercube sampling (LHS) (McKay et al. 1979) .  ... 
arXiv:2206.12194v1 fatcat:gp6zlclz6zbe3doeevuvcppcci

Bayesian Design of Experiments for Nonlinear Dynamic System Identification

Susanne Zaglauer
2012 Proceedings of the Fifth International Conference on Simulation Tools and Techniques  
In order to defuse this critique a new online Bayesian design for the nonlinear dynamic system identification is introduced, which serves the flexibility and which is concurrently more resistant against  ...  In extreme cases the boundary points of the experimental space are the experimental candidates.  ...  The Latin Hypercube Design In this schema only one sampling point is in every column and row of a grid.  ... 
doi:10.4108/icst.simutools.2012.247734 dblp:conf/simutools/Zaglauer12 fatcat:rcjdvyn7zjdj7ihfs5g6vnnq2a

Special Issue on Uncertainty Quantification in Multiscale System Design and Simulation

Yan Wang, Laura Swiler
2017 ASCE-ASME J of Risk & Uncertainty in Engineering SystemsPart B: Mechanical Engineering  
In the paper by Li and Mahadevan entitled "Sensitivity Analysis of a Bayesian Network," global sensitivity analysis is applied to evaluate the correlation between variables in Bayesian networks.  ...  This paper explored the parameters of the modified embedded-atom method through a design of experiments and Latin hypercube sampling approach to better understand how individual modified embedded-atom  ... 
doi:10.1115/1.4037447 fatcat:d4wx3apmtrfx7icw47mpj6uji4

Turboelectric Uncertainty Quantification and Error Estimation in Numerical Modelling

Mosab Alrashed, Theoklis Nikolaidis, Pericles Pilidis, Soheil Jafari
2020 Applied Sciences  
Some of the current models, like Monte Carlo and Latin hypercube sampling, are reliable.  ...  The results show that the electrical elements in turboelectric systems can have decent outcomes in statistical analysis.  ...  It is possible to integrate a Latin hypercube sampling technique into a Monte Carlo model and operate it Sampling points Sampling points Latin Hypercube The Latin hypercube algorithm involves a stratified  ... 
doi:10.3390/app10051805 fatcat:ui6vkfxwrndf5lndlpxnl7uvje

Bayesian inverse modeling of the atmospheric transport and emissions of a controlled tracer release from a nuclear power plant

Donald D. Lucas, Matthew Simpson, Philip Cameron-Smith, Ronald L. Baskett
2017 Atmospheric Chemistry and Physics  
model inputs that affect the transport and dispersion of a trace gas released from a coastal California nuclear power plant are quantified using ensemble simulations, machine-learning algorithms, and Bayesian  ...  The WRF output is used to drive tens of thousands of FLEXPART dispersion simulations that sample a uniform distribution of six emissions inputs.  ...  Additional discussion of Latin hypercube sampling for ensemble modeling is given in Lucas et al. (2013) .  ... 
doi:10.5194/acp-17-13521-2017 fatcat:7fkw72hqrbfnhbopkh5bdse2s4
« Previous Showing results 1 — 15 out of 2,023 results