1,720,999 research outputs found
wedeling/deep_active_subspace_data: Final revision of DAS article
No full textThe is the final version meant to reproduce the results of:
W.N. Edeling, "On the deep active subspace method", SIAM/ASA Journal on Uncertainty Quantification, 2022
wedeling/MD-active-subspace: final release
No full text<p>Replicate results of: Wouter Edeling, Maxime Vassaux, Yiming Yang, Shunzhou Wan, Serge Guillas, Peter Coveney, Global ranking of the sensitivity of interaction potential contributions within classical molecular dynamics force fields, NPJ computational materials, 2024.</p>
wedeling/deep_active_subspace_data: Release with DAS article
No full textData and source codes needed to reproduce the results of:
W.N. Edeling, On the deep active subspace method (submitted), 2021
wedeling/deep_active_subspace_data
No full textMaterials accompanying W.N. Edeling, "On the deep active subspace method", SIAM/ASA Journal on Uncertainty Quantification, 2022
EasySurrogate - phys_D branch
No full textThis branch of EasySurrogate contains all software required to reproduce the results from:
D. Crommelin, W. Edeling, "Resampling with neural networks for stochastic parameterization in multiscale systems
FabCovidsim
No full textThis is a FabSim3 / EasyVVUQ plugin for Covid-19 simulation. It was used to compute the ensembles of the following paper:
Edeling, Wouter and Hamid, Arabnejad and Sinclair, Robert and Suleimenova, Diana and Gopalakrishnan, Krishnakumar and Bosak, Bartosz and Groen, Derek and Mahmood, Imran and Crommelin, Daan and Coveney, Peter, The Impact of Uncertainty on Predictions of the CovidSim Epidemiological Code, 2020
On the deep active-subspace method
Full text linkThe deep active-subspace method is a neural-network based tool for the propagation of uncertainty through computational models with high-dimensional input spaces. Unlike the original active-subspace method, it does not require access to the gradient of the model. It relies on an orthogonal projection matrix constructed with Gram–Schmidt orthogonalization to reduce the input dimensionality. This matrix is incorporated into a neural network as the weight matrix of the first hidden layer (acting as an orthogonal encoder), and optimized using back propagation to identify the active subspace of the input. We propose several theoretical extensions, starting with a new analytic relation for the derivatives of Gram–Schmidt vectors, which are required for back propagation. We also study the use of vector-valued model outputs, which is difficult in the case of the original active-subspace method. Additionally, we investigate an alternative neural network with an encoder without embedded orthonormality, which shows equally good performance compared to the deep active-subspace method. Two epidemiological models are considered as applications, where one requires supercomputer access to generate the training data
wedeling/Gram_Schmidt_Derivatives: Release with DAS article
No full textContains a Python script to compute the derivatives of Gram Schmidt vectors, as well as a symbolic math Jupyter notebook to validate the expressions. The mathematical derivation of the derivative expressions can also be found
EasySurrogate
No full textThis is the M42 release of EasySurrogate, a toolkit designed to facilitate the creation of surrogate models for multiscale simulations. The development of this software is funded by the EU Horizon 2020 Verified Exascale Computing for Multiscale Applications (VECMA) project
- …
