How to plot network performance?
显示 更早的评论
I am trying to plot the network performance using following code
[net2,tr] = trainNetwork(augimdsTrain,lgraph,options);
figure
plot(tr.TrainingLoss,'b-')
hold on
x = 1 : length(tr.ValidationLoss);
y = tr.ValidationLoss;
idx = ~any(isnan(y),1);
plot(x(idx),y(idx),'--k','Marker','.','MarkerSize' ,12);
plotperf(tr)
but got the following error
Unrecognized field name "num_epochs".
Error in plotperform>update_plot (line 258)
numEpochs = tr.num_epochs;
Error in plotperform (line 109)
plotData = update_plot(param,fig,plotData,update_args{:});
my tr variable have
>> tr
tr =
struct with fields:
TrainingLoss: [1×9187 double]
TrainingAccuracy: [1×9187 double]
ValidationLoss: [1×9187 double]
ValidationAccuracy: [1×9187 double]
BaseLearnRate: [1×9187 double]
FinalValidationLoss: 0.0322
FinalValidationAccuracy: 99.6214
回答（2 个）
yanqi liu
2022-3-7
0 个投票
PLOTPERF Plot network performance.
the tr get by train not trainNetwork，so may be should plot perf by self，such as use plotroc(targets,outputs)
2 个评论
john karli
2022-3-7
yanqi liu
2022-3-7
targets：real label for test data
outputs：predict label for test data
Mahesh Taparia
2022-3-9
0 个投票
Hi
You can select the 'Plots' name value pair of trainingOptions function to 'training-progress' and then train the network. You will be able to see the training plot, and you can export and save that. For more information, you can refer this documentation.
Hope it helps!
类别
在 帮助中心 和 File Exchange 中查找有关 Deep Learning Toolbox 的更多信息
2022-3-4
2022-3-9
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
