Accelerating SGD with momentum for over-parameterized learning Original paper: https://openreview.net/forum?id=r1gixp4FPH