Comment on Why are deep neural networks hard to train?parentComments−benanne11yI haven't had any luck so far with rmsprop, adagrad and adadelta. SGD + Nesterov momentum has served me best.
Comments
I haven't had any luck so far with rmsprop, adagrad and adadelta. SGD + Nesterov momentum has served me best.