Leveraging inductive bias of neural networks for learning without explicit human annotations

Open in new window