MATLAB: Neural network programming error.

Deep Learning Toolboxneural networktrain

 clear all;clc
 a=rand(1,1000);
 b=rand(1,1000);
 c=rand(1,1000);
 y=a*2+b*3+c*5
 I=[a; b; c;]
 T=y
 net = newff([0 1;0 1 ;0 1],[10 1]);
 net=train(net,I,T);
 J1=sim(net,I);
 Op=sim(net,[1 1 1]');
As i am testing system with inputs [1 1 1] then the output must be neat to 10, but it gives 1. what is wrong with this code?

Best Answer

 clear all; clc
 rng('default')    % Initialized the RNG
 a = rand(1,1000);
 b = rand(1,1000);
 c = rand(1,1000);
 y = a*2+b*3+c*5;  % Added semicolon


 I = [a; b; c;];   % Added semicolon
 T = y;            % Added semicolon
 net = newff([0 1;0 1 ;0 1],[10 1]);
 net=train(net,I,T);
 J1=sim(net,I);
 MSE = mse(J1-T)
 Op=sim(net,[1 1 1]')
 %  Warning: NEWFF used in an obsolete
% way. 
% > In obs_use at 18
%   In newff>create_network at 127
%   In newff at 102
%   In Untitledgh at 9 
%   See help for NEWFF to update calls to 
%   the   new argument list.
 % MSE =   19.6749
 % Op =     1
 net = newff( I ,T ,10 );  % Current version
 net=train(net,I,T);
 J1=sim(net,I);
 MSE = mse(J1-T)       
 Op=sim(net,[1 1 1]')   
 % MSE =   4.4650e-009
 % Op  =     9.9969
To be sure the obsolete version contains a bug,
 Ntrials = 15
 for i=1:Ntrials
    net = newff([0 1;0 1 ;0 1],[10 1]);
    net=train(net,I,T);
    J1=sim(net,I);
    MSE(i) = mse(J1-T);
    Op(i)=sim(net,[1 1 1]');
 end
 result = [ MSE' OP' ]
%  result =
% 
%    19.6749    1.0000












%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.7560    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
%    22.1988    1.0000
%    19.6749    1.0000
%    19.6749    1.0000
Hope this helps
Thank you for formally accepting my answer
Greg
Related Question