Why do I have the same validation accuracy for every epoch?

Question

Ioana Cretu 2021-5-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/824215-why-do-i-have-the-same-validation-accuracy-for-every-epoch

评论： Ioana Cretu 2021-5-18

I developed a CNN for ECG arrhythmia classification and when I train the model I obtain the same validation accuracy for all of the 50 epoch. Can you please tell me what is wrong? I tried to modify the parameters, also the structure, but the validation accuracy is unchanged (80.1%).

Labels=cnnLabels(labels); %Divide and add LABELS in classes from 1 to 5 
[heartbeats, ~, ~] = featureNormalize(heartbeats); % Normalization
%% Spliting the data in training and testing sets
PercentNumFiles =round(0.90*length(heartbeats)); %90% of the files for training and 10% for testing
trainingPercent=round(0.9*PercentNumFiles);%90% of the training set is used for training and the other 10 for training validation
randomNum(:,1)=randperm(length(heartbeats)); %random selection of the files 
Xtrain=heartbeats(randomNum(1:trainingPercent),:);
Xvalidation=heartbeats(randomNum(trainingPercent+1:PercentNumFiles),:);
Xtest=heartbeats(randomNum(PercentNumFiles+1:end),:);
Ytrain=Labels(randomNum(1:trainingPercent));
Yvalidation=Labels(randomNum(trainingPercent+1:PercentNumFiles));
Ytest=Labels(randomNum(PercentNumFiles+1:end));
%% CNN 
clear ECG fs ind before after anntype cleanECG PercentNumFiles randomNum
height = 1;
width = 300;
channels = 1;
Xtrain = reshape(Xtrain,[height, width, channels, length(Xtrain)]);
Xvalidation=reshape(Xvalidation,[height, width, channels, length(Xvalidation)]);
Xtest = reshape(Xtest,[height, width, channels, length(Xtest)]);
Ytrain=categorical(Ytrain);
Yvalidation=categorical(Yvalidation);
Ytest=categorical(Ytest);
%% CNN construction
% classes = [1 2 3 4 5];
% classWeights = [0.1 0.7 0.6 0.9 0.3];
classWeights = 1./countcats(Ytrain);
classWeights = classWeights'/mean(classWeights);
Layers=[
    imageInputLayer([height,width,channels]); %'DataAugmentation', 'none'); %'Normalization', 'none');
    
    convolution2dLayer([1 3], 256,'stride',[1 1], 'padding','same');  %Filter window size = [1 5], No of filters = 64, stride = [1 1];
    
    convolution2dLayer([1 3], 256,'stride',[1 1], 'padding','same');  
    reluLayer();
    dropoutLayer();
    maxPooling2dLayer([1 2],'stride',[1 2]); %PoolSize = [1 2], Stride = [1 1] 
    
    convolution2dLayer([1 3], 128,'stride',[1 1], 'padding','same');  
    reluLayer();
    
    convolution2dLayer([1 3], 128,'stride',[1 1], 'padding','same');  
    reluLayer();
    
    convolution2dLayer([1 3], 64,'stride',[1 1], 'padding','same');  
    reluLayer();
    dropoutLayer();
    maxPooling2dLayer([1 2],'stride',[1 2]); %PoolSize = [1 2], Stride = [1 1]
    
    
    fullyConnectedLayer(256); 
    dropoutLayer();
    fullyConnectedLayer(128);
    fullyConnectedLayer(5); %Reduce to five output classes
    softmaxLayer();
    classificationLayer();
    ];
%% Options of training
options = trainingOptions('sgdm','InitialLearnRate',0.001,'MaxEpochs',50, ...
    'LearnRateSchedule','piecewise', ...
    'LearnRateDropFactor',0.1, ...
    'LearnRateDropPeriod', 3,...
    'L2Regularizatio',1.0000e-04, ...
    'MiniBatchSize', 60,...
    'ValidationData',{Xvalidation, Yvalidation},...
    'Plots','training-progress');
convnet = trainNetwork(Xtrain,Ytrain,Layers,options);

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。

请先登录，再回答此问题。

Answer 1

Aditya Patil 2021-5-10

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/824215-why-do-i-have-the-same-validation-accuracy-for-every-epoch#answer_696210

As per my understanding, the data you have in unidimensional, and is time variant.

It might be a better option to consider this as a sequence classification problem. Have a look at this example for classifying ECG using LSTMs. There are also video tutorials for the same.

For the current model, it might be predicting the largest class for all the observations. You can check this by looking at the confusion matrix or the outputs.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ioana Cretu 2021-5-18

在 MATLAB Online 中打开

Thank you! I really wanted to use 1D CNN or at least to combine it with LSTMs. I modified the code so I can use the CNN features for LSTM but I have another error

Layers=[
    sequenceInputLayer(inputSize,'Normalization', 'zscore', 'Name','input'); 
    sequenceFoldingLayer('Name','fold')
    
    convolution2dLayer([1 7], 16,'stride',[1 1], 'padding','same','Name','conv1')
    batchNormalizationLayer('Name','bn1')
    maxPooling2dLayer([1 2],'stride',[1 2],'Name','mpool1')
    
    convolution2dLayer([1 7], 32,'stride',[1 1], 'padding','same','Name','conv2')
    batchNormalizationLayer('Name','bn2')
    reluLayer('Name','relu1')
    maxPooling2dLayer([1 2],'stride',[1 2],'Name','mpool2')
    
    convolution2dLayer([1 5], 64,'stride',[1 1], 'padding','same','Name','conv3') 
    batchNormalizationLayer('Name','bn3')
    reluLayer('Name','relu2')
    
    convolution2dLayer([1 5], 128,'stride',[1 1], 'padding','same','Name','conv4')
    batchNormalizationLayer('Name','bn4')
    reluLayer('Name','relu3')
    
    convolution2dLayer([1 3], 256,'stride',[1 1], 'padding','same','Name','conv5')  
    batchNormalizationLayer('Name','bn5')
    reluLayer('Name','relu4')
    maxPooling2dLayer([1 2],'stride',[1 2],'Name','mpool3')
    
    convolution2dLayer([1 3], 512,'stride',[1 1], 'padding','same','Name','conv6')
    batchNormalizationLayer('Name','bn6')
    reluLayer('Name','relu5')
    maxPooling2dLayer([1 2],'stride',[1 2],'Name','mpool4')
%     dropoutLayer('Name','dropout')
    
    sequenceUnfoldingLayer('Name','unfold')
    flattenLayer('Name','flatten')
    
    bilstmLayer(200,'Name','lstm')
    
    reluLayer('Name','relu6')
    fullyConnectedLayer(256,'Name','fc1')
    reluLayer('Name','relu7')
    fullyConnectedLayer(128,'Name','fc2')
    reluLayer('Name','relu8')
    fullyConnectedLayer(5,'Name','fc3')
    softmaxLayer('Name','softmax')
    classificationLayer('Name','classification')
    ];
lgraph = layerGraph(Layers);
lgraph = connectLayers(lgraph,'fold/miniBatchSize','unfold/miniBatchSize');

When I run this

convnet = trainNetwork(Xtrain,Ytrain,Layers,options);

I have this error

Error using trainNetwork (line 170)
Invalid network.
Caused by:
    Layer 'fold': Unconnected output. Each layer output must be connected to the input of another layer.
    Detected unconnected outputs:
        output 'miniBatchSize'
    Layer 'unfold': Unconnected input. Each layer input must be connected to the output of another layer.
    Detected unconnected inputs:
        input 'miniBatchSize'
        