Learning Compact Hash Codes for Multimodal Representations Using Orthogonal Deep Structure | Synapse