Add support for unsupervised models

[nshaud]

Maybe out of scope for this toolbox.

The toolbox currently work with the assumption that the models are supervised. Working with unsupervised models (e.g. autoencoders) could be helpful.

• Support other criterions in models.py (easy)
• Support other metrics in metrics() (medium)
• Either support no target or target = data (e.g. autoencoder) (hard)

[mengxue-rs]

Interesting features, I'd like to try although I am not familiar with coding unsupervised models. -)