/homomorphismvae

Primary LanguagePythonApache License 2.0Apache-2.0

Symmetry Based Representation Learning through Homomorphism AutoEncoders

Homomorphism AutoEncoder HAE is a model trained on observed transitions (o_t, g_t, o_{t+1}) to jointly learn a group representation of the observed actions g_t and a representation of the observations o.

Main scripts are provided in "./displacementae/homomorphism/".

Run commands are provided in the "./displacementae/homomorphism/README.rst".