Label Efficient Learning of Transferable Representations acrosss Domains and Tasks

We propose a framework that learns a representation transferable across different domains and tasks in a data efficient manner.