This repository contains an implmentation of a deep Gaussian inversion using Galerkin method.
This repository is tested to run against these packages
Numba
version should be later than0.47
,CUPY
version7.0.0bc3
, withCUDA
version10.1
,- Standard installation of
numpy
andscipy
, h5py
.- Python version
3.7
. - Replace your
.conda/envs/xxxx/lib/python3.7/site-packages/cupy/linalg/norms.py
with the one in thepatch
folder.
Explanation of files and folders:
- To run the MCMC, run
twoDsim.py
. - This repository is a companion of an arXiv paper, Non-Stationary Multi-layered Gaussian Priors for Bayesian Inversion
As this repository is a working folder, things might break without notification. Should you face any problem, please contact muhammad.emzir@aalto.fi.