Model Interpretability and Rationale Extraction by Input Mask Optimization

Marc Brinner, Sina Zarriess

Concurrent to the rapid progress in the development of neural-network based models in areas like natural language processing and computer vision, the need for creating explanations…