Multi-Variance Replica Exchange SGMCMC for Inverse and Forward Problems Via Bayesian PINN

SSRN Electronic Journal（2021）

引用 4|浏览0

暂无评分

摘要

•A multi-variance replica exchange Langevin dynamics MCMC algorithm is developed to solve PDE inverse problems.•The method is able to discover the multi-mode of the problem and save the computational cost.•The algorithm can be used to optimize the Bayesian physics informed neural networks.•An unbiased particles swapping rate is proposed and a convergence theorem is developed.

查看译文

关键词

Replica exchange,Bayesian physics-informed neural network,Inverse problem,Machine learning,Deep learning,Neural network

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要