Skip to content

greedy nonlinear autoregression for multifidelity modeling

Notifications You must be signed in to change notification settings

wayXing/GreedyNAR

Repository files navigation

GreedyNAR

Greedy nonlinear autoregression for multifidelity modeling. (Multi-fidelity Gaussian process regression with active learning based on a greedy approach)

MATLAB code for the paper, Greedy nonlinear autoregression for multifidelity computer models at different scales (https://www.sciencedirect.com/science/article/pii/S2666546820300124).

This repository provides a modification of the nonlinear autoregression method for a model sequential construction based on greedy approaches. It also provides the stochastic collocation method for multifidelity modeling and Gaussian process with sequential learning for comparison.

Please refer Demo_synthe_01.m and Demo_synthe_02.m for the usage of the code. Please add all subfolder to your path to start using the code.

The following figures (generated by running Demo_synthe_02.m) show the performance of greedyNAR and stochastic collocation for synthetic three-fidelity data.

Normal Gaussian process with only high-fidelity observations: Normal Gaussian process with only limited high-fidelity observations

Stochastic collocation without low-fidelity observations: Stochastic collocation without low-fidelity observations

Stochastic collocation with low-fidelity observations: Stochastic collocation with low-fidelity observations

GreedyNAR without low-fidelity observations: GreedyNAR without low-fidelity observations

GreedyNAR with low-fidelity observations: GreedyNAR with low-fidelity observations

About

greedy nonlinear autoregression for multifidelity modeling

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages