Bidirectional Multi-Modal Margin Ranking Criterion is widely adopted for multimodal similarity
learning.

[1] A. Karpathy and L. Fei-Fei. Deep visual-semantic alignments for generating image descriptions, In CVPR, 2015.
[2] A. Karpathy, A. Joulin, and F. F. F. Li. Deep fragment embeddings for bidirectional image sentence mapping. In NIPS, 2014.
[3] L. Wang, Y. Li, and S. Lazebnik. Learning deep structurepreserving image-text embeddings. In CVPR, 2016.

@takatosp1 Could you add documentation and unit test to this PR?