Create trainDAEs.py

[jamessergeant]

Python script to perform all fine-tune training using DAEs.

• Parallel training of MMAE and SMAE.
• If multiple RBMs available and unspecified, request user to select.
• Perform checks for specified RBMs.
• Perform checks for availability of datasets, provide link to download if they aren't available.
• Save trained DAEs differently named from provided DAEs (unless they aren't going to be provided).
• Save trained DAEs differently named from other trained DAEs.
• Add trained DAEs to appropriate lists for use in operation.
• Can the fprop function be deployed? Can try pickling the theano function
• Make executable

[jamessergeant]

There might be a way to use model.getstate and model.setstate however need to determine how to save the theano graph.