1. Understanding the difficulty of training deep feedforward neural networks;glorot;Proc 13th Int Conf Artif Intell Statist,2010
2. Self-normalizing neural networks;klambauer;Proc 31st Int Conf Neural Inf Process Syst,2017
3. A Sufficient Condition for Convergences of Adam and RMSProp
4. On the convergence of adam and beyond;reddi;arXiv 1904 09237,2019