@becker

Dropout: a simple way to prevent neural networks from overfitting.

, , , , and . Journal of machine learning research, 15 (1): 1929--1958 (2014)