.. index:: todo
OptimizedClassifier: to automatically select the model in an easy and non-biased(i.e. non-cheating) way, so it could become a part of any
more advanced pipeline as a regular classifier
Add ability to add/modify custom attributes to a dataset instance
IPython mode
Possibly make NiftiDataset default to float32 when it sees that the data are ints
Along with ICA mapper, we should add a PLS mapper