Performance errors for neural networks

1 次查看（过去 30 天）

显示更早的评论

Fabrizio Bernardi 2020-10-29

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/630498-performance-errors-for-neural-networks

关闭： MATLAB Answer Bot 2021-8-20

在 MATLAB Online 中打开

Hello eveyone, I saw that in shallow ANN, after training a network, I can see Training,Validation and testing performances using

plotperf(tr);

However, in my case I created a regression output layer to have a customized Loss function, and I used for training progress something like

options = trainingOptions('adam', ...
    'MaxEpochs',50, ...
    'GradientThreshold',2, ...
    'InitialLearnRate',0.003, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropPeriod',33, ...
    'LearnRateDropFactor',0.75,...
    'Verbose',1, ...
    'MiniBatchSize',50, ...
    'Plots','training-progress');
net = trainNetwork(myInput,myTarget,layers,options);

Is it possible from this to have some functions that allows me to have a similiar perfomance plot as in the case of plotperform? Or a struct like tr?

Thank you for your help!

2 个评论
显示无隐藏无

Aditya Patil 2020-11-17

What issues did you face with 'Plots','training-progress' argument?

Fabrizio Bernardi 2020-11-17

I wanted to include validation as well, but I only now have seen the option to includit through XValidation and Yvalidation data! All solved, thanks for your time!

此问题已关闭。