Improving zero-shot learning by mitigating the hubness problem

BibTex: