A fully differentiable pore network for digital reconstruction of porous media

This work highlights a fully differentiable framework for constructing pore networks from porosimetry and anisotropic permeability data. The resulting network can be scaled using a trained Gaussian Process model.