How does matlab solve the samples in neural networks?

1 Ansicht (letzte 30 Tage)
Huang Huang
Huang Huang am 17 Dez. 2019
Beantwortet: Huang Huang am 19 Dez. 2019
Usualy the column represents the sample number.But when I test the following codes, the results(weight1 and weight2) are different.Who would help me? Thank you very much!
%------------
clear
clc
X=[0 0 1;0 1 1;1 0 1;1 1 1];
D=[0; 0; 1; 1];
X=X';
D=D';
net=network;
net.numInputs=1;
net.numLayers=1;
net.OutputConnect=1;
net.inputConnect(1,1)=1;
net.trainFcn='traingd';
net.trainParam.epochs=1;
net1=train(net,X,D);
weight1=net1.IW{1,1}
net2=net;
for k=1:4
net2=train(net2,X(:,k),D(k));
end
weight2=net2.IW{1,1}

Akzeptierte Antwort

Huang Huang
Huang Huang am 19 Dez. 2019
Now,I know why!
In matlab,'traingd' indicates the batch GD method. From above codes,net1 is the batch algorithm,while net2 is the equivalent GD algorithm.
So the difference is right.

Weitere Antworten (0)

Community Treasure Hunt

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

Start Hunting!

Translated by