Label Efficient Audio Classification Through Multitask Learning And Self Supervision

While deep learning has been incredibly successful in modeling tasks with large, carefully curated labeled datasets, its application to problems with limited labeled data remains a challenge.