Statistical maximum-entropy models for turbulence generation
Abstract: Models of stationary (homogeneous) processes may be defined as maximum-entropy distributions conditioned by a family of statistical moments that are invariant to translations. A major challenge is to prescribe the moments such that the resulting model is able to generate similar geometric patterns such as tourbillons in Turbulence. We introduce phase harmonic moments to capture these crucial non-Gaussian properties of stationary processes. The phase harmonics are inspired from the rectifier non-linearity in deep neural networks. An extension of the model for inertial particle distribution driven by Turbulent winds will also be presented. This is a joint-work with A. Brochard, B. Błaszczyszyn and S. Mallat. Some turbulence data are provided by K. Matsuda, T. Oujia and K. Schneider.
Lien visio : https://lpensl.my.webex.com/
-> salle : Webconf3