Towards Theoretically Understanding Why S GD
  Generalizes Better Than A DAM in Deep Learning

   Pan Zhou , Jiashi Feng , Chao Ma , Caiming Xiong , Steven HOI , Weinan E
   
     Salesforce Research,  National University of Singapore,  Princeton University
  {pzhou,shoi,cxiong}@salesforce.com elefjia@nus.edu.sg {chaom@, weinan@math.}princeton.edu

                       Abstract
     It is not clear yet why A DAM-alike adaptive gradient algorithms suffer from worse
     generalization ...