To address these challenges, this letter proposes an approach based on representation learning and the novel intermediate-state informed loss function modification (ISIL). ISIL leverages unlabeled ...

Specifically, meta-learning is executed over a pretrained network for few-shot classification. Furthermore, self-supervised learning is exploited to fit the target domain distribution by training on ...