Autoencoder TODOs

[gdkrmr]

TODOs for the Autoencoder:

• write tests for the other activation function types
• better stopping criteria
• more than one training step, continue training a dimRedResult object
• handle the installation of tensorflow and keras (this should be handled by the user -> IGNORE)

[topepo]

(sorry, I should have put my other comments in your repo)

Can you name the resulting variable differently? They aren't really related to PCA. Maybe use the prefix "AE" or similar?

[gdkrmr]

• I did the renaming, autoencoders are called nl-PCA in literature related to Meteorology, but autoencoder is probably the better name.
• I did a quick search and I am not so happy with using Keras, mainly, because it seems that it does not allow the use of shared weights, there are autoencoders where the decoder weights consist of the transposed encoder weights and a couple of other fancy things that seem difficult or impossible in Keras.
• On the upside I think users could supply their own graph in Keras which would make something like convolutional autoencoders possible without me having to implement every special case and serialization seems super easy.