The example of Train Network Using Federated Learning that is given in MathWorks documentation is not working

Question

debojit sharma 2023-12-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2063982-the-example-of-train-network-using-federated-learning-that-is-given-in-mathworks-documentation-is-no

回答： Harsha Vardhan 2024-1-5

I am trying to run the example of Train Network Using Federated Learning given in the MathWorks documentation ( https://in.mathworks.com/help/deeplearning/ug/train-network-using-federated-learning.html ). But I am getting the following error:

Though it showing that Undefined function 'preprocessMiniBatch' . I have included the following function of 'preprocessMiniBatch' as given in the Mathworks documentation

Everytime I run the code I am getting this error. I am unable to understand where I am making a mistake. I am using MATLAB 2023a version, CPU and 8GB RAM. I am looking for a solution since a month now. Someone please help me in solving this problem. I will be grateful to you.

2 个评论
显示无隐藏无

Walter Roberson 2023-12-25

If you are not already doing so, try putting preprocessMiniBatch into its own .m file

debojit sharma 2023-12-28

Thank you @Walter Roberson sir. Now the program is working. Sir, I want to plot the confusion matrix of the training phase and test phase. How to plot the confusion matrix. Please suggest some solution.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Harsha Vardhan 2024-1-5

在 MATLAB Online 中打开

Hi Debojit Sharma,

I understand that you faced an error while using the Federated Learning example. It appears that you were able to resolve this issue following a comment from community. Later, you wanted to plot a confusion matrix for this example.

Confusion Matrix can be plotted using the ‘ confusionmat’ function. Please check the relevant documentation here - https://www.mathworks.com/help/stats/confusionmat.html

To integrate confusion matrix computation for training and testing phases into your existing federated learning code, you can collect predictions and actual labels from the global model for both training and testing datasets and then use these to compute the confusion matrices. There are other possible ways of calculaing the confusion matrix too.

You can check the code mofifications below.

Just like datastores were created for test and validation data, create a datastore for training data as below.

fileList = [];
labelList = [];
for i = 1:numWorkers
    tmp = imdsTestVal{i};
    
    fileList = cat(1,fileList,tmp.Files);
    labelList = cat(1,labelList,tmp.Labels);    
end
imdsGlobalTestVal = imageDatastore(fileList);
imdsGlobalTestVal.Labels = labelList;
[imdsGlobalTest,imdsGlobalVal] = splitEachLabel(imdsGlobalTestVal,0.5,"randomized");
augimdsGlobalTest = augmentedImageDatastore(inputSize(1:2),imdsGlobalTest);
augimdsGlobalVal = augmentedImageDatastore(inputSize(1:2),imdsGlobalVal);
%% Code for creating a datastore for training data
fileList = [];
labelList = [];
for i = 1:numWorkers
    tmp = imdsTrain{i};
    
    fileList = cat(1,fileList,tmp.Files);
    labelList = cat(1,labelList,tmp.Labels);    
end
imdsGlobalTrainVal = imageDatastore(fileList);
imdsGlobalTrainVal.Labels = labelList;
augimdsGlobalTrain = augmentedImageDatastore(inputSize(1:2),imdsGlobalTrainVal);

Similarly, create a 'minibatchqueue' object for training data.

mbqGlobalVal = minibatchqueue(augimdsGlobalVal, ...
    MiniBatchSize=miniBatchSize, ...
    MiniBatchFcn=preProcess, ...
    MiniBatchFormat=["SSCB",""]);
%Code for creating a minibatchqueue for training data
mbqGlobalTrain = minibatchqueue(augimdsGlobalTrain, ...
    MiniBatchSize=miniBatchSize, ...
    MiniBatchFcn=preProcess, ...
    MiniBatchFormat=["SSCB",""]);

After calculating the accuracy, you can plot the confusion matrices for all the training and testing data as below.

accuracy = computeAccuracy(globalModel,mbqGlobalTest,classes);
%Code for displaying training confusion matrix
trainConfusionMat = createConfusionMatrix(globalModel, mbqGlobalTrain, classes);
figure;
confusionchart(trainConfusionMat);
title('Training Confusion Matrix');
%Code for displaying testing confusion matrix
testConfusionMat = createConfusionMatrix(globalModel, mbqGlobalTest, classes);
figure;
confusionchart(testConfusionMat);
title('Testing Confusion Matrix');

The below function creates a confusion mattrix using the 'confusionmat' MATLAB function.

%function for calculating Confusion Matrix
function confusionMat = createConfusionMatrix(net, mbq, classes)
    allYPred = [];
    allTTest = [];
    
    shuffle(mbq);
    while hasdata(mbq)
        [XTest, TTest] = next(mbq);
        TTest = onehotdecode(TTest, classes, 1)';
        
        YPred = predict(net, XTest);
        YPred = onehotdecode(YPred, classes, 1)';
        
        allTTest = [allTTest; TTest];
        allYPred = [allYPred; YPred];
    end
    
    confusionMat = confusionmat(categorical(allTTest), categorical(allYPred));
end

Hope this helps in resolving your query!

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

The example of Train Network Using Federated Learning that is given in MathWorks documentation is not working

2 个评论
显示无隐藏无

回答（1 个）

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

The example of Train Network Using Federated Learning that is given in MathWorks documentation is not working

2 个评论 显示 无隐藏 无

回答（1 个）

0 个评论 显示 -2更早的评论隐藏 -2更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

2 个评论
显示无隐藏无

0 个评论
显示 -2更早的评论隐藏 -2更早的评论