Here is another paper which shows that one can use PyTorch to create a simulation environment to train (learn) an inhomogenous medium to guide a solution of the

Here is another paper which shows that one can use PyTorch to create a simulation environment to train (learn) an inhomogenous medium to guide a solution of the wave equation such that the mapping between input and output closely fits a dataset, in this case classifying spoken vowels. https://advances.sciencemag.org/content/5/12/eaay6946 https://advances.sciencemag.org/content/5/12/eaay6946 And the repo: https://github.com/fancompute/wavetorch https://github.com/fancompute/wavetorch