Princeton University Star Formation/ISM Rendezvous (SFIR)

The ubiquitous filamentary patterns that are observed in the interstellar medium (ISM) are the result of the interplay between turbulence, magnetic fields and gravity. These patterns are examples of the highly non-Gaussian structures that emerge naturally from non-linear physical processes. As our understanding of the magnetized ISM largely relies on the comparison of observations with numerical MHD simulations and phenomenological models we need an adequate statistical description of non-Gaussian, filamentary structures. Such a description would provide a way to quantify the impact of the physical parameters on the statistical properties of the emerging structures.In this talk, I will present the Reduced Wavelet Scattering Transform (RWST), a low variance statistical description of non-Gaussian physical processes. This representation has been obtained through a reduction of the Wavelet Scattering Transform (WST) of Mallat et al., by distributing the information it contains into a few number of terms whose physical meaning is identified. It does not need any training stage to be computed, and allows to perform in-depth physical studies as well as realistic syntheses of ISM fields with very few coefficients (less than 100 in our studies).