If you want help, give more information. Like me old value and the new value. This can make a big difference in training.
Fred Le 8 déc. 2016 07:41, "Edward Wang" <[email protected]> a écrit : > The net is not changed > I change the learning rate and the size of layers > > > > 在 2016年12月7日星期三 UTC+8下午11:19:27，Adepu Ravi Sankar写道： >> >> >> Could you please point to the modification you have done. >> >> On Wednesday, December 7, 2016 at 2:28:10 PM UTC+5:30, Edward Wang wrote: >>> >>> Hello all: >>> >>> I an a freshman to Theano and machine learning. >>> >>> I found a code of multilayer preceptron and modified it. >>> The code as well as the input and output data is attached, >>> >>> I am facing with a question that the convergent rate of this code is too >>> slow. >>> My input is 4096 samples with 39 parameters and the dimension of my >>> output is 1. >>> >>> There is not satisfying results (the square error is less than 10^-5 or >>> much lower) even after 300 iterations. >>> I add the momentum and change the learning rate for every iteration, >>> however, it seems doesn't work. >>> >>> Could you please help me to check this code and give me some guidance? >>> >>> Thank you very much! >>> >>> >>> -- > > --- > You received this message because you are subscribed to the Google Groups > "theano-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
