Published April 15, 2015 | Version v1
Dataset Open

Improving zero-shot learning by mitigating the hubness problem

  • 1. University of Trento


Data associated with the paper Improving zero-shot learning by mitigating the hubness problem, proceedings of ICLR 2015 (International Conference on Learning Representations), workshop track.

Abstract: The zero-shot paradigm exploits vector-based word representations extracted from text corpora with unsupervised methods to learn general mapping functions from other feature spaces onto word space, where the words associated to the nearest neighbours of the mapped vectors are used as their linguistic labels. We show that the neighbourhoods of the mapped elements are strongly polluted by hubs, vectors that tend to be near a high proportion of items, pushing their correct labels down the neighbour list. After illustrating the problem empirically, we propose a simple method to correct it by taking the proximity distribution of potential neighbours across many mapped vectors into account. We show that this correction leads to consistent improvements in realistic zero-shot experiments in the cross-lingual, image labeling and image retrieval domains.


Files (568.7 MB)

Name Size Download all
568.7 MB Preview Download

Additional details


COMPOSES – Compositional Operations in Semantic Space 283554
European Commission