How to learn from scarce data? How to handle underrepresented classes? How to learn from smaller datasets?
How to take advantage of multi-view data? How to leverage multi task and distributed learning?
How to go beyond the supervised learning paradigm? Learn with unlabeled or weakly labelled data?
: How to best represent audio data specificities in deep learning models? And improve interpretability?
How to generate new data from interpretable models learnt from few data with good generalization ability?