probabilistic neural network coding problem
1 visualización (últimos 30 días)
Mostrar comentarios más antiguos
abdulhamid
el 16 de Dic. de 2013
Respondida: Greg Heath
el 17 de Dic. de 2013
i have create this simple code and trying to make classification between two inputs, but, after run the code the output is not overlapping the input and is not fitting the target, even the output it self i am not sure is looks like ANN output? or PNN has different output functions.
any one can give suggestion please.
regards
code is here:
clc
format short
p=[1.7572 8.3747 6.6941 4.0114 1.7572 1.9934 4.9658 8.4807 5.3349 5.5587 2.7213 1.9934 8.8202 4.9658 5.754 7.809 4.3451 4.7672
1.30E+08 5.81E+05 -6.63E+05 3.47E+06 1.30E+08 1.19E+08 2.68E+08 -2.39E+05 4.99E+05 -9.24E+04 -5.09E+06 1.19E+08 -7.06E+04 2.68E+08 -3.67E+05 -2.80E+05 -8.74E+05 -1.77E+05
-0.3045 -0.4317 -0.8981 0.1139 -0.3045 -0.0431 1.6113 -0.5305 -0.4704 -0.0686 -0.0464 -0.431 -0.4884 1.6113 -0.4654 -0.1465 -0.6266 0.6488
5.02E+07 2.09E+07 7.13E+06 2.48E+07 5.02E+07 4.92E+07 1.71E+08 7.47E+06 7.14E+06 7.85E+06 5.50E+07 4.92E+07 7.06E+06 1.71E+08 7.56E+06 2.96E+07 1.14E+06 1.17E+07
];
t=[0 1 0 1 0 0 1 0 0 0 1 0 0 1 0 1 0 0
];
plottools('on');
spread= 10;
net=newpnn(p,t,spread);
net.performFcn;
toc;
tic;
plottools('on');
hold on;
%net.trainparam.show=1;
nettrainparam.epochs=1000;
%net.trainparam.goal=0.0002;
%net.trainParam.searchFcn = 'srchbac';
nettrainparam.lr=0.5;
%net=train(net,p,t);
y=sim(net,p);
plotregression(t,y);
%plottraining (t,y);
plottools('on');
error=abs(y-t);
plot(error);
%y=sim(net,p);
grid on;
legend off;
plottools('on');
%plot(p,t,'*');
%e = t-y
% perf = mse(e);
% perf = sse(e);
End
0 comentarios
Respuesta aceptada
Greg Heath
el 17 de Dic. de 2013
1. Class 1 has 2 duplicate points and class 2 has 1 duplicate point
2. There is an 8 orders of magnitude difference between variables (1,3) and variables (2,4). Therefore, I standardized the data to have zero mean and unit variance.
3. Using newpnn with 0 < spread < 2 I was not able to separate the classes.
I recommend you use newrb and loop over spread values to obtain the best result.
Thank you for formally accepting my answer
Greg
0 comentarios
Más respuestas (0)
Ver también
Categorías
Más información sobre Define Shallow Neural Network Architectures en Help Center y File Exchange.
Productos
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!