/DNM

Deep Neural Maps

Primary LanguagePythonMIT LicenseMIT

Theano implementation of the Deep Neural Maps. Methods of DNM:

  • get_locations(X): returns the locations of the mapping on the lattice
  • get_embeddings(X): returns the embeddings of X
  • get_reconstructed(X): returns the output of the AE for X
  • decode_embeddings(z)
  • backproject_map(): returns the numpy array of the DNM map
  • get_bmu_array(X): returns the BMU map element of X

Learning progress on mnist:

./map_animation_mnist.gif

./backprojections/back_projected_mnist.png

./backprojections/back_projected_coil.png

./backprojections/back_projected_fashion.png

./backprojections/back_projected_celebA.png