0

It's been 3 days since i'm trying to train many neural networks to predict sin(x) function, i'm using matlab 2016b (i have to work with it in my assignement)

what i did :

  • change layers
  • duplicate dataset (big , small)
  • add/sub periods
  • shuffle the data
  • change neural's number per layer
  • change learning function
  • change the transfer function and mapped the target

all that with no good prediction, can anyone explain me what i'm doing wrong , and it would be very helpful to paste any good book for ("preparing dataset befor traing", "knowing the best NN's structure for your project",... and any book seems helpful)

my actual code : (i'm using nntool for the training )

%% input and target
input = 0:pi/100:8*pi;
target = sin(input) ;


plot(input,sin(input)),
hold on,

inputA = input;
targetA = target;
plot(inputA,targetA),
hold on,
%simulate input
output=sim(network2,inputA);
plot(inputA,output,'or')
hold off

0 Answers0