Deep Structured Prediction with Nonlinear Output Transformations Colin Graber Ofer Meshi Alexander Schwing

While deep nets are a very convenient tool these days, enabling rapid progress in both industry and academia, their training is known to require significant amounts of data.