One of the common problems of neural networks, especially those with many layers, consists of their lengthy training time. We attempt to solve this problem at the algorithmic level, proposing a simple parallel design which ...