Cross-view Action Recognition via Transferable Dictionary Learning.