manifold learning

Autoencoders for stochastic dynamics get geometric regularization

A new method adds tangent-bundle and inverse-consistency regularization to autoencoders for reduced stochastic simulators.