Matlab neural network, how to force the use of certain sets for training, validation and testing?

前端 未结 1 1794
[愿得一人]
[愿得一人] 2021-01-24 12:26

If you use the GUI nnstart, you provide only input and output, but you can\'t decide which rows are going to be training, validation and test set because they are chosen randoml

1条回答
  •  旧时难觅i
    2021-01-24 12:49

    You can use any of the GUI's launched from NNSTART to create some sample training code. Then customize that code by setting the following data division values:

    net.divideFcn = 'divideind'; % Divide data by indices (i.e. not randomly)
    net.divideParam.trainInd = [... training indices you want ...];
    net.divideParam.valInd = [... training indices you want ...];
    net.divideParam.testInd = [... training indices you want ...];
    

    0 讨论(0)
提交回复
热议问题