How can I make 'Feedforwardnet' behave like 'patternnet'?

2 次查看（过去 30 天）

oneclock 2016-11-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/312273-how-can-i-make-feedforwardnet-behave-like-patternnet

编辑： oneclock 2016-11-15

I don't know how to make 'Feedforwardnet' behave like 'patternnet'. This is my codes.

[inputs,targets] = cancer_dataset;
% Create a Pattern Recognition Network
hiddenLayerSize = 10;
net = feedforwardnet(hiddenLayerSize);
% Set up Division of Data for Training, Validation, Testing
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;
% Train the Network
[net,tr] = train(net,inputs,targets);
% Test the Network
outputs = net(inputs);

This is the example of the patternnet without using feedforwardnet. I see the two network's output is different: patternnet produce the output of 0~1 data.

So, I want to add code so that 'feedforwardnet' behaves like 'patternnet' without using transferfunction'mapminmax'.

I think that the difference between 'feedforwardnet' and 'patternnet' is the added output at the network: Sigmoid hidden and softmax, is this correct? If this is correct, I want to add code to 'feedforwardnet'.

Thanks.