Neural network multi fidelity methods for uncertainty quantification in kinetic theory
Please login to view abstract download link
In this talk, we will discuss the construction of novel multi-fidelity methods for kinetic equation where the low fidelity surrogates are composed by both simplfied macroscopic models and by neural network representation of the full model. Thanks to control variate approaches we show how these techniques are capable to reduce the variance of standard Monte Carlo techniques.
