S2SCAT: Differentiable and accelerated spherical scattering transforms

Web Github

S2SCAT is a Python package for computing scattering covariances on the sphere using JAX. It exploits autodiff to provide differentiable transforms, which are also deployable on hardware accelerators (e.g. GPUs and TPUs), leveraging the differentiable and accelerated spherical harmonic and wavelet transforms implemented in S2FFT and S2WAV, respectively. Scattering covariances are useful both for field-level generative modelling of complex non-Gaussian textures and for statistical compression of high dimensional field-level data, a key step of e.g. simulation based inference.

code

Publications

Generative models of astrophysical fields with scattering transforms on the sphere

L. Mousset, E. Allys, M. A. Price, J. Aumont, J. -M. Delouis, L. Montier, J. D. McEwen

Preprint PDF Code

S2SCAT: Differentiable and accelerated spherical scattering transforms

Publications

Talks