Code not running and showing busy

1 次查看（过去 30 天）
Rajdeep Ghosh
Rajdeep Ghosh 2020-8-3
评论： Rajdeep Ghosh 2020-8-8
Dear programmers
I have written a piece of code. Please check after providing convergence criteria with while loop the code is not running and only showing as busy. Please help.
Regards
Rajdeep
input= xlsread('Input');
output = xlsread('Output');
bias = ones(1,3)*0.0005; %bias values
a=-1;
b=1;
rng('default')
V= a+(b-a)*rand(6,5); %weights between the input(6 nodes) and hidden(5 nodes) layers
rng('default')
W = a+(b-a)*rand(5,1); %weights between the hidden(5 nodes) and output(1 node) layers
transposed_input = input'; %inputs
d = output'; %target output
%initialization of del[W] and del[V]
del_W1=0;
del_W2=0;
del_W3=0;
del_W4=0;
del_W5=0;
del_V1=0;
del_V2=0;
del_V3=0;
del_V4=0;
del_V5=0;
del_V6=0;
del_V7=0;
del_V8=0;
del_V9=0;
del_V10=0;
del_V11=0;
del_V12=0;
del_V13=0;
del_V14=0;
del_V15=0;
del_V16=0;
del_V17=0;
del_V18=0;
del_V19=0;
del_V20=0;
del_V21=0;
del_V22=0;
del_V23=0;
del_V24=0;
del_V25=0;
del_V26=0;
del_V27=0;
del_V28=0;
del_V29=0;
del_V30=0;
for epoch=1:6000 %number of iterations
alpha=0.1; %learning rate
Mu=0.1; %Momentum Constant
N=80; %Number of training cases
for k=1:N
%Feed-Forward Network starts
%Output to input neurons
OI(1,k)=1./(1+exp(-(transposed_input(1,k)+bias(1,1))));
OI(2,k)=1./(1+exp(-(transposed_input(2,k)+bias(1,1))));
OI(3,k)=1./(1+exp(-(transposed_input(3,k)+bias(1,1))));
OI(4,k)=1./(1+exp(-(transposed_input(4,k)+bias(1,1))));
OI(5,k)=1./(1+exp(-(transposed_input(5,k)+bias(1,1))));
OI(6,k)=1./(1+exp(-(transposed_input(6,k)+bias(1,1))));
%Input to hidden neurons
IH1(1,k)= (OI(1,k)*V(1,1))+(OI(2,k)*V(2,1))+ (OI(3,k)*V(3,1))+ (OI(4,k)*V(4,1))+ (OI(5,k)*V(5,1))+(OI(6,k)*V(6,1))+bias(1,2);
IH2(1,k)= (OI(1,k)*V(1,2))+(OI(2,k)*V(2,2))+ (OI(3,k)*V(3,2))+ (OI(4,k)*V(4,2))+ (OI(5,k)*V(5,2))+(OI(6,k)*V(6,2))+bias(1,2);
IH3(1,k)= (OI(1,k)*V(1,3))+(OI(2,k)*V(2,3))+ (OI(3,k)*V(3,3))+ (OI(4,k)*V(4,3))+ (OI(5,k)*V(5,3))+(OI(6,k)*V(6,3))+bias(1,2);
IH4(1,k)= (OI(1,k)*V(1,4))+(OI(2,k)*V(2,4))+ (OI(3,k)*V(3,4))+ (OI(4,k)*V(4,4))+ (OI(5,k)*V(5,4))+(OI(6,k)*V(6,4))+bias(1,2);
IH5(1,k)= (OI(1,k)*V(1,5))+(OI(2,k)*V(2,5))+ (OI(3,k)*V(3,5))+ (OI(4,k)*V(4,5))+ (OI(5,k)*V(5,5))+(OI(6,k)*V(6,5))+bias(1,2);
%Output to hidden neurons
OH1(1,k) = (exp(IH1(1,k)+bias(1,2)) - exp(-(IH1(1,k)+bias(1,2))))./(exp(IH1(1,k)+bias(1,2)) + exp(-(IH1(1,k)+bias(1,2))));
OH2(1,k) = (exp(IH2(1,k)+bias(1,2)) - exp(-(IH2(1,k)+bias(1,2))))./(exp(IH2(1,k)+bias(1,2)) + exp(-(IH2(1,k)+bias(1,2))));
OH3(1,k) = (exp(IH3(1,k)+bias(1,2)) - exp(-(IH3(1,k)+bias(1,2))))./(exp(IH3(1,k)+bias(1,2)) + exp(-(IH3(1,k)+bias(1,2))));
OH4(1,k) = (exp(IH4(1,k)+bias(1,2)) - exp(-(IH4(1,k)+bias(1,2))))./(exp(IH4(1,k)+bias(1,2)) + exp(-(IH4(1,k)+bias(1,2))));
OH5(1,k) = (exp(IH5(1,k)+bias(1,2)) - exp(-(IH5(1,k)+bias(1,2))))./(exp(IH5(1,k)+bias(1,2)) + exp(-(IH5(1,k)+bias(1,2))));
%Input to Output neuron
IO(1,k)=OH1(1,k)*W(1,1)+OH2(1,k)*W(2,1)+OH3(1,k)*W(3,1)+OH4(1,k)*W(4,1)+OH5(1,k)*W(5,1)+bias(1,3);
%Output to Output neuron
Out(1,k)=IO(1,k)+bias(1,3);
%forward step calculation corresponding to each training case of a batch run terminates here
%BackPropagation of [V] and [W] starts
%Finding Mean Squared Error(E)
error(1,k)=d(1,k)-Out(1,k);
diff_Out(1,k)=[error(1,k).^2];
E= sum(diff_Out(1,k))./(2*N);
end
while (E >= 0.0004) %convergence criteria
for p=1:N
delta(1,p)=Out(1,p)*(1-Out(1,p))*error(1,p);
del=sum(delta(1,p));
%summation of Output to hidden neurons for 'k' training cases
OHS1=sum(OH1(1,p));
OHS2=sum(OH2(1,p));
OHS3=sum(OH3(1,p));
OHS4=sum(OH4(1,p));
OHS5=sum(OH5(1,p));
%summation of Output to input neurons for 'k' training cases
OI1=sum(OI(1,p));
OI2=sum(OI(2,p));
OI3=sum(OI(3,p));
OI4=sum(OI(4,p));
OI5=sum(OI(5,p));
OI6=sum(OI(6,p));
end
delta2_1=OHS1*(1-OHS1)*W(1,1)*del;
delta2_2=OHS2*(1-OHS2)*W(2,1)*del;
delta2_3=OHS3*(1-OHS3)*W(3,1)*del;
delta2_4=OHS4*(1-OHS4)*W(4,1)*del;
delta2_5=OHS5*(1-OHS5)*W(5,1)*del;
%delta[W]
del_W1=-alpha*OHS1*del+Mu*del_W1;
del_W2=-alpha*OHS2*del+Mu*del_W2;
del_W3=-alpha*OHS3*del+Mu*del_W3;
del_W4=-alpha*OHS4*del+Mu*del_W4;
del_W5=-alpha*OHS5*del+Mu*del_W5;
%delta[V]
del_V1=-alpha*OI1*delta2_1+Mu*del_V1;
del_V2=-alpha*OI2*delta2_1+Mu*del_V2;
del_V3=-alpha*OI3*delta2_1+Mu*del_V3;
del_V4=-alpha*OI4*delta2_1+Mu*del_V4;
del_V5=-alpha*OI5*delta2_1+Mu*del_V5;
del_V6=-alpha*OI6*delta2_1+Mu*del_V6;
del_V7=-alpha*OI1*delta2_2+Mu*del_V7;
del_V8=-alpha*OI2*delta2_2+Mu*del_V8;
del_V9=-alpha*OI3*delta2_2+Mu*del_V9;
del_V10=-alpha*OI4*delta2_2+Mu*del_V10;
del_V11=-alpha*OI5*delta2_2+Mu*del_V11;
del_V12=-alpha*OI6*delta2_2+Mu*del_V12;
del_V13=-alpha*OI1*delta2_3+Mu*del_V13;
del_V14=-alpha*OI2*delta2_3+Mu*del_V14;
del_V15=-alpha*OI3*delta2_3+Mu*del_V15;
del_V16=-alpha*OI4*delta2_3+Mu*del_V16;
del_V17=-alpha*OI5*delta2_3+Mu*del_V17;
del_V18=-alpha*OI6*delta2_3+Mu*del_V18;
del_V19=-alpha*OI1*delta2_4+Mu*del_V19;
del_V20=-alpha*OI2*delta2_4+Mu*del_V20;
del_V21=-alpha*OI3*delta2_4+Mu*del_V21;
del_V22=-alpha*OI4*delta2_4+Mu*del_V22;
del_V23=-alpha*OI5*delta2_4+Mu*del_V23;
del_V24=-alpha*OI6*delta2_4+Mu*del_V24;
del_V25=-alpha*OI1*delta2_5+Mu*del_V25;
del_V26=-alpha*OI2*delta2_5+Mu*del_V26;
del_V27=-alpha*OI3*delta2_5+Mu*del_V27;
del_V28=-alpha*OI4*delta2_5+Mu*del_V28;
del_V29=-alpha*OI5*delta2_5+Mu*del_V29;
del_V30=-alpha*OI6*delta2_5+Mu*del_V30;
end
%updated [W] and [V] matrix
W(1,1)=W(1,1)+del_W1;
W(2,1)=W(2,1)+del_W2;
W(3,1)=W(3,1)+del_W3;
W(4,1)=W(4,1)+del_W4;
W(5,1)=W(5,1)+del_W5;
V(1,1)= V(1,1)+del_V1;
V(2,1)= V(2,1)+del_V2;
V(3,1)= V(3,1)+del_V3;
V(4,1)= V(4,1)+del_V4;
V(5,1)= V(5,1)+del_V5;
V(6,1)= V(6,1)+del_V6;
V(1,2)= V(1,2)+del_V7;
V(2,2)= V(2,2)+del_V8;
V(3,2)= V(3,2)+del_V9;
V(4,2)= V(4,2)+del_V10;
V(5,2)= V(5,2)+del_V11;
V(6,2)= V(6,2)+del_V12;
V(1,3)= V(1,3)+del_V13;
V(2,3)= V(2,3)+del_V14;
V(3,3)= V(3,3)+del_V15;
V(4,3)= V(4,3)+del_V16;
V(5,3)= V(5,3)+del_V17;
V(6,3)= V(6,3)+del_V18;
V(1,4)= V(1,4)+del_V19;
V(2,4)= V(2,4)+del_V20;
V(3,4)= V(3,4)+del_V21;
V(4,4)= V(4,4)+del_V22;
V(5,4)= V(5,4)+del_V23;
V(6,4)= V(6,4)+del_V24;
V(1,5)= V(1,5)+del_V25;
V(2,5)= V(2,5)+del_V26;
V(3,5)= V(3,5)+del_V27;
V(4,5)= V(4,5)+del_V28;
V(5,5)= V(5,5)+del_V29;
V(6,5)= V(6,5)+del_V30;
end

请先登录，再回答此问题。

采纳的回答

Raunak Gupta
Raunak Gupta 2020-8-8
Hi Rajdeep,
From the code I see you are using first for loop inside the main for loop for updating the value of E. And the while loop condition is based on E only. So, you are not updating the variable on which the while loop is dependent inside the while loop, the problem it will create is let’s say the value of E from first for loop is >= 0.0004, then the while loop will not terminate because its value is not changing at all in that loop or any subsequent loop.
I think there is something missing in the implementation about updating E within the while loop. You may check again what are equation mathematically for writing the backpropagation and the termination condition.
Hope it helps!
  1 个评论
Rajdeep Ghosh
Rajdeep Ghosh 2020-8-8
Thank you Raunak for indicating me with the implementation of the code.
You are absolutely right. I changed the while into if-end statement and now its working fine. I was stuck in the algorithm. Thanks again for indicating the error. The error E will be updated in each epoch and not in while loop.

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

标签

产品


版本

R2017a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by