Follow
Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult. (arXiv:2310.17087v1 [cs.LG]) http://arxiv.org/abs/2310.17087
Good regularity creates large learning rate implicit biases: edge of stability, balancing, and catapult. (arXiv:2310.17087v1 [cs.LG]) http://arxiv.org/abs/2310.17087