Batch normalization provably avoids ranks collapse for randomly initialised deep networks

Hadi Daneshmand, Jonas Moritz Kohler, Francis R. Bach, Thomas Hofmann, Aurélien Lucchi. Batch normalization provably avoids ranks collapse for randomly initialised deep networks. In Hugo Larochelle, Marc'Aurelio Ranzato, Raia Hadsell, Maria-Florina Balcan, Hsuan-Tien Lin, editors, Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6-12, 2020, virtual. 2020. [doi]

Abstract

Abstract is missing.