Unsupervised Few-Shot Audio Classification

In few-shot unsupervised classification, we assume that at the model pre-training stage, only unlabelled data is available. This contrasts the typical case where we assume labelled data is available for pre-training.