The output size (100) of the last layer does not match theresponse size (40000).
1 次查看(过去 30 天)
显示 更早的评论
I designed a very simple dnn, whose structure is shown in the figure.
layers = [
%Input Layers
featureInputLayer(40000);
%Hidden Layers
fullyConnectedLayer(256)
reluLayer
fullyConnectedLayer(256)
reluLayer
fullyConnectedLayer(256)
reluLayer
% Output Layers
fullyConnectedLayer(100)
regressionLayer];
And the code i used to generate my trainset —— 2D signals and 2D signals affected by noise, I have put this .m file as an attachment. I want to be able to use neural networks to remove noise from the signal.
When I input my training set to train it, the system prompts me that there is an error.
layers = [
%Input Layers
featureInputLayer(40000);
%Hidden Layers
fullyConnectedLayer(256)
reluLayer
fullyConnectedLayer(256)
reluLayer
fullyConnectedLayer(256)
reluLayer
% Output Layers
fullyConnectedLayer(100)
regressionLayer];
options = trainingOptions('adam', ...
'InitialLearnRate',0.001,...
'LearnRateSchedule','piecewise',...
'LearnRateDropFactor',0.9,...
'LearnRateDropPeriod',40,...
'ExecutionEnvironment', 'cpu', ...
'MiniBatchSize', 32, ...
'MaxEpochs', 500, ...
'ValidationData',{VaSet_o,VaSet_n},...
'L2Regularization',1.0000e-05,...
'Shuffle', 'every-epoch', ...
'Plots', 'training-progress', ...
'Verbose', false);
net = trainNetwork(TrainSet_n, TrainSet_o, layers, options);
Error using trainNetwork
Invalid training data. The output size (100) of the last layer does not match theresponse size (40000).
I am quite sure that the response size of my training set is 100 instead of 40000
OK, then I change the output size to 40000 but didn't modify my train set.
And the system the system told me that another error occurred.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1298780/image.png)
The output size (40000) of the last layer does not match the response size (100).
Now how does the response size change back to 100????
WHY??????????
I really hope someone can help me, please.
This is very important to me.
2 个评论
Matt J
2023-2-17
编辑:Matt J
2023-2-17
It does seem strange, but please post your code as formatted text, e.g.,
a=1;
b=2;
instead of as images, so it is easy for us to copy/paste and run it. Also, please attach a subset of your features and response training data in a .mat file so that we can use it to run the code.
采纳的回答
更多回答(0 个)
另请参阅
类别
在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!