Chasing the Long Tail: What Neural Networks Memorize and Why

One of our favorite talks at the Simons Institute this past month was by Vitaly Feldman (Apple). He spoke about the utility of memorization in deep learning algorithms, and touched on some intriguing tensions between data privacy and fairness. 

His presentation was part of the workshop on Societal Considerations and Applications, within the research program on Data-Driven Decision Processes.