How to Train 1d CNN on Custom dataset in matrix form in MATLAB

Hi everyone, i hope you are doing well.

In my previous question : https://www.mathworks.com/matlabcentral/answers/1649260-how-to-train-cnn-on-custom-dataset-in-matrix-form

yanqi liu answer the question with 2D CNN, But i wanted to train 1D CNN

i have the following dataset myFile.txt includes 102x5,in which first 4 coloums are the Number of Observation and the last column are the Discrete labels/Classes for the dataset. I want to train 1D-CNN on this dataset

sz = size(dataset);
dataset = dataset(randperm(sz(1)),:);
traindata=dataset(:,1:4);
trainlabel=categorical(dataset(:,5));
classes = unique(trainlabel)
numClasses = numel(unique(trainlabel))
PD = 0.80 ;  
Ptrain = []; Ttrain = [];
Ptest = []; Ttest = [];
for i = 1 : length(classes)
    indi = find(trainlabel==classes(i));
    indi = indi(randperm(length(indi)));
    indj = round(length(indi)*PD);
    Ptrain = [Ptrain; traindata(indi(1:indj),:)]; Ttrain = [Ttrain; trainlabel(indi(1:indj),:)]; 
    Ptest = [Ptest; traindata(indi(1+indj:end),:)]; Ttest = [Ttest; trainlabel(indi(1+indj:end),:)]; 
end
Ptrain=(reshape(Ptrain', [4,1,1,size(Ptrain,1)]));
Ptest=(reshape(Ptest', [4,1,1,size(Ptest,1)]));
layers = [imageInputLayer([4 1 1])
    convolution2dLayer([3 1],3,'Stride',1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
    dropoutLayer
    fullyConnectedLayer(numClasses)
    softmaxLayer
    classificationLayer];
options = trainingOptions('adam', ...
    'MaxEpochs',3000, ...
    'Shuffle','every-epoch', ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'ValidationData',{Ptest,Ttest},...
    'ExecutionEnvironment', 'cpu', ...
    'ValidationPatience',Inf);
net = trainNetwork(Ptrain,Ttrain,layers,options);

[yanqi liu]

yes，sir，may be “convolution1dLayer” is not exist，may be use convolution2dLayer and set one dimension to 1 to use it as convolution1dLayer

[Med Future]

@yanqi liu https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.convolution1dlayer.html

here is the 1D CNN layer

[Med Future]

@yanqi liu where i can change in convolution2dLayer to set dimension 1

Answer 1

[yanqi liu]

yes，sir，if 2021b has convolution1dLayer，so we can make the cnn as follows，then we can try train it

layers = [sequenceInputLayer(4)
    convolution1dLayer(3,32,Padding="causal")
    reluLayer
    globalMaxPooling1dLayer
    dropoutLayer
    fullyConnectedLayer(5)
    softmaxLayer
    classificationLayer];
layers

layers = 
  8×1 Layer array with layers:

     1   ''   Sequence Input           Sequence input with 4 dimensions
     2   ''   Convolution              32 3 convolutions with stride 1 and padding 'causal'
     3   ''   ReLU                     ReLU
     4   ''   1-D Global Max Pooling   1-D global max pooling
     5   ''   Dropout                  50% dropout
     6   ''   Fully Connected          5 fully connected layer
     7   ''   Softmax                  softmax
     8   ''   Classification Output    crossentropyex

[Med Future]

@yanqi liu have you tried it?

[yanqi liu]

dataset = [0.008,2.5699e+09,-60.397,5.0922e-06,1
0.0084998,2.5698e+09,-59.954,5.0719e-06,1
0.0089996,2.5701e+09,-59.638,5.0947e-06,1
0.0094995,2.5696e+09,-59.246,5.0906e-06,1
0.0099995,2.5701e+09,-58.858,5.0083e-06,1
0.010499,2.5696e+09,-58.538,4.9623e-06,1
0.010999,2.5702e+09,-58.004,4.96e-06,1
0.011499,2.5705e+09,-57.557,4.99e-06,1
0.03,2.5802e+09,-58.372,9.9782e-06,2
0.03055,2.5803e+09,-58.253,9.9844e-06,2
0.0311,2.5801e+09,-58.22,9.9826e-06,2
0.03165,2.5799e+09,-58.096,1.0018e-05,2
0.094902,2.5798e+09,-48.896,1.0022e-05,2
0.095451,2.5805e+09,-48.955,1.0003e-05,2
0.096001,2.5804e+09,-48.721,1.0002e-05,2
0.096551,2.5796e+09,-48.758,1.0011e-05,2
0.097101,2.5799e+09,-48.639,1.0021e-05,2
0.097652,2.5797e+09,-48.812,1.002e-05,2
0.098202,2.5797e+09,-48.622,1.0005e-05,2
0.098752,2.5801e+09,-48.592,1.0009e-05,2
0.099302,2.5796e+09,-48.571,9.989e-06,2
0.099852,2.5804e+09,-48.707,1.0001e-05,2
0.1004,2.5799e+09,-48.712,1.0013e-05,2
0.10095,2.5797e+09,-48.67,1.0008e-05,2
0.1015,2.5804e+09,-48.555,1.0002e-05,2
0.1,2.5904e+09,-56.358,1.4981e-05,3
0.10057,2.59e+09,-55.861,1.4994e-05,3
0.10114,2.5902e+09,-55.625,1.4918e-05,3
0.10171,2.5902e+09,-55.158,1.5037e-05,3
0.10228,2.5901e+09,-54.942,1.5006e-05,3
0.10285,2.5904e+09,-54.477,1.5099e-05,3
0.10342,2.5899e+09,-54.001,1.5055e-05,3
0.10399,2.5896e+09,-53.86,1.5047e-05,3
0.10456,2.5901e+09,-53.599,1.4958e-05,3
0.10513,2.5902e+09,-53.212,1.5004e-05,3
0.1057,2.5896e+09,-52.693,1.5076e-05,3
0.10627,2.59e+09,-52.356,1.4937e-05,3
0.10684,2.5903e+09,-52.188,1.4938e-05,3
0.10741,2.5902e+09,-51.769,1.4918e-05,3
0.10798,2.5897e+09,-51.339,1.4932e-05,3
0.13,2.6001e+09,-58.171,4.9896e-05,4
0.1461,2.6004e+09,-53.86,5.0058e-05,4
0.14725,2.5996e+09,-53.42,4.9917e-05,4
0.1484,2.6002e+09,-53.152,5.0033e-05,4
0.14955,2.6e+09,-52.855,5.0033e-05,4
0.1507,2.5996e+09,-52.568,5.0067e-05,4
0.15185,2.6002e+09,-52.298,5.0028e-05,4
0.153,2.5999e+09,-52.044,5.0002e-05,4
0.15415,2.5996e+09,-51.865,4.992e-05,4
0.1553,2.5998e+09,-51.591,4.997e-05,4
0.15645,2.5996e+09,-51.336,5.0093e-05,4
0.1576,2.6002e+09,-51.19,5.0095e-05,4
0.15875,2.6005e+09,-50.951,4.9887e-05,4
0.1599,2.5995e+09,-50.693,4.9926e-05,4
0.16105,2.6e+09,-50.461,5.012e-05,4
0.1622,2.6004e+09,-50.302,5.0023e-05,4
0.16335,2.6001e+09,-50.048,4.9884e-05,4
0.1645,2.5997e+09,-49.878,5.0005e-05,4
0.16565,2.5997e+09,-49.791,5.0064e-05,4
0.1668,2.6e+09,-49.524,5.0041e-05,4
0.16795,2.6005e+09,-49.388,5.0036e-05,4
0.1691,2.6004e+09,-49.174,5.0053e-05,4
0.17025,2.5995e+09,-48.996,5.0016e-05,4
0.1714,2.5996e+09,-48.934,4.9977e-05,4
0.17255,2.6e+09,-48.782,5.0048e-05,4
0.1737,2.6005e+09,-48.727,5.0117e-05,4
0.17485,2.5997e+09,-48.641,4.9965e-05,4
0.176,2.6003e+09,-48.484,4.9884e-05,4
0.17715,2.5998e+09,-48.39,4.989e-05,4
0.1783,2.6003e+09,-48.42,4.9925e-05,4
0.17945,2.6005e+09,-48.393,4.9896e-05,4
0.1806,2.5997e+09,-48.316,4.9955e-05,4
0.18175,2.6e+09,-48.215,4.99e-05,4
0.1829,2.5997e+09,-48.125,5.0049e-05,4
0.18405,2.5997e+09,-48.28,4.9908e-05,4
0.1852,2.6004e+09,-48.275,5.0074e-05,4
0.18635,2.6001e+09,-48.135,5.0007e-05,4
0.1875,2.6004e+09,-48.314,4.9951e-05,4
0.18865,2.6e+09,-48.185,4.9916e-05,4
0.1898,2.5995e+09,-48.271,5.0097e-05,4
0.19095,2.5996e+09,-48.395,4.9928e-05,4
0.1921,2.6004e+09,-48.396,4.9988e-05,4
0.19325,2.5997e+09,-48.455,5.0102e-05,4
0.1944,2.6e+09,-48.575,5.0096e-05,4
0.19555,2.5998e+09,-48.663,5.0079e-05,4
0.25,2.6037e+09,-75.123,2.0454e-06,5
0.25075,2.6042e+09,-73.315,1.999e-06,5
0.2515,2.6044e+09,-71.653,1.9823e-06,5
0.25225,2.6038e+09,-69.994,2.0639e-06,5
0.253,2.604e+09,-68.391,1.9963e-06,5
0.25375,2.6043e+09,-67.257,1.9299e-06,5
0.2545,2.6039e+09,-66.164,2.0163e-06,5
0.25525,2.6036e+09,-65.639,1.9588e-06,5
0.256,2.6038e+09,-65.321,1.913e-06,5
0.25675,2.6045e+09,-65.28,2.0615e-06,5
0.2575,2.6039e+09,-65.538,1.9969e-06,5
0.25825,2.6041e+09,-66.246,1.9181e-06,5
0.259,2.6041e+09,-67.172,2.0295e-06,5
0.25975,2.6041e+09,-68.554,2.0292e-06,5
0.2605,2.6043e+09,-69.869,2.0366e-06,5
0.26125,2.6043e+09,-71.51,1.9323e-06,5
0.262,2.6036e+09,-73.437,1.9219e-06,5];

sz = size(dataset);
dataset = dataset(randperm(sz(1)),:);
traindata=dataset(:,1:4);
trainlabel=categorical(dataset(:,5));

classes = unique(trainlabel);
numClasses = numel(unique(trainlabel));

PD = 0.80 ;  
Ptrain = []; Ttrain = [];
Ptest = []; Ttest = [];
for i = 1 : length(classes)
    indi = find(trainlabel==classes(i));
    indi = indi(randperm(length(indi)));
    indj = round(length(indi)*PD);
    Ptrain = [Ptrain; traindata(indi(1:indj),:)]; Ttrain = [Ttrain; trainlabel(indi(1:indj),:)]; 
    Ptest = [Ptest; traindata(indi(1+indj:end),:)]; Ttest = [Ttest; trainlabel(indi(1+indj:end),:)]; 
end
Ptrain2 = []; Ttrain2 = [];
Ptest2 = []; Ttest2 = [];
Ttrain2=categorical(Ttrain);
Ttest2=categorical(Ttest);
for i = 1 : size(Ptrain,1)
    Ptrain2{i} = Ptrain(i,:)';
    %Ttrain2{i} = Ttrain(i);
end
for i = 1 : size(Ptest,1)
    Ptest2{i} = Ptest(i,:)';
    %Ttest2{i} = Ttest(i);
end

layers = [sequenceInputLayer(4)
    convolution1dLayer(3,32,Padding="causal")
    reluLayer
    globalMaxPooling1dLayer
    dropoutLayer
    fullyConnectedLayer(5)
    softmaxLayer
    classificationLayer];

options = trainingOptions('adam', ...
    'MaxEpochs',3000, ...
    'Shuffle','every-epoch', ...
    'Verbose',false, ...
    'ValidationData',{Ptest2,Ttest2},...
    'ExecutionEnvironment', 'auto', ...
    'ValidationPatience',Inf);
[net,dinfo] = trainNetwork(Ptrain2,Ttrain2,layers,options)

net = 
  SeriesNetwork with properties:

         Layers: [8×1 nnet.cnn.layer.Layer]
     InputNames: {'sequenceinput'}
    OutputNames: {'classoutput'}

dinfo = struct with fields:
               TrainingLoss: [13.8038 13.2205 14.1926 14.5815 14.5815 12.6373 13.8038 12.2484 13.2205 14.7759 14.3870 13.2205 13.9982 12.6373 12.6373 13.2205 13.9982 13.8038 13.9982 13.6094 13.0261 13.2205 13.2205 12.8317 13.9982 12.0540 12.8317 … ]
           TrainingAccuracy: [13.4146 17.0732 10.9756 8.5366 8.5366 20.7317 13.4146 23.1707 17.0732 7.3171 9.7561 17.0732 12.1951 20.7317 20.7317 17.0732 12.1951 13.4146 12.1951 14.6341 18.2927 17.0732 17.0732 19.5122 12.1951 24.3902 19.5122 13.4146 … ]
             ValidationLoss: [13.5510 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN 8.7683 NaN NaN NaN … ]
         ValidationAccuracy: [15 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN 45 NaN NaN NaN NaN NaN … ]
              BaseLearnRate: [0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 … ]
        FinalValidationLoss: 13.5510
    FinalValidationAccuracy: 15
     OutputNetworkIteration: 3000

[Med Future]

@yanqi liu why validation accuracy is so low?

[Med Future]

@yanqi liu code is not working

[yanqi liu]

dataset = [0.008,2.5699e+09,-60.397,5.0922e-06,1
0.0084998,2.5698e+09,-59.954,5.0719e-06,1
0.0089996,2.5701e+09,-59.638,5.0947e-06,1
0.0094995,2.5696e+09,-59.246,5.0906e-06,1
0.0099995,2.5701e+09,-58.858,5.0083e-06,1
0.010499,2.5696e+09,-58.538,4.9623e-06,1
0.010999,2.5702e+09,-58.004,4.96e-06,1
0.011499,2.5705e+09,-57.557,4.99e-06,1
0.03,2.5802e+09,-58.372,9.9782e-06,2
0.03055,2.5803e+09,-58.253,9.9844e-06,2
0.0311,2.5801e+09,-58.22,9.9826e-06,2
0.03165,2.5799e+09,-58.096,1.0018e-05,2
0.094902,2.5798e+09,-48.896,1.0022e-05,2
0.095451,2.5805e+09,-48.955,1.0003e-05,2
0.096001,2.5804e+09,-48.721,1.0002e-05,2
0.096551,2.5796e+09,-48.758,1.0011e-05,2
0.097101,2.5799e+09,-48.639,1.0021e-05,2
0.097652,2.5797e+09,-48.812,1.002e-05,2
0.098202,2.5797e+09,-48.622,1.0005e-05,2
0.098752,2.5801e+09,-48.592,1.0009e-05,2
0.099302,2.5796e+09,-48.571,9.989e-06,2
0.099852,2.5804e+09,-48.707,1.0001e-05,2
0.1004,2.5799e+09,-48.712,1.0013e-05,2
0.10095,2.5797e+09,-48.67,1.0008e-05,2
0.1015,2.5804e+09,-48.555,1.0002e-05,2
0.1,2.5904e+09,-56.358,1.4981e-05,3
0.10057,2.59e+09,-55.861,1.4994e-05,3
0.10114,2.5902e+09,-55.625,1.4918e-05,3
0.10171,2.5902e+09,-55.158,1.5037e-05,3
0.10228,2.5901e+09,-54.942,1.5006e-05,3
0.10285,2.5904e+09,-54.477,1.5099e-05,3
0.10342,2.5899e+09,-54.001,1.5055e-05,3
0.10399,2.5896e+09,-53.86,1.5047e-05,3
0.10456,2.5901e+09,-53.599,1.4958e-05,3
0.10513,2.5902e+09,-53.212,1.5004e-05,3
0.1057,2.5896e+09,-52.693,1.5076e-05,3
0.10627,2.59e+09,-52.356,1.4937e-05,3
0.10684,2.5903e+09,-52.188,1.4938e-05,3
0.10741,2.5902e+09,-51.769,1.4918e-05,3
0.10798,2.5897e+09,-51.339,1.4932e-05,3
0.13,2.6001e+09,-58.171,4.9896e-05,4
0.1461,2.6004e+09,-53.86,5.0058e-05,4
0.14725,2.5996e+09,-53.42,4.9917e-05,4
0.1484,2.6002e+09,-53.152,5.0033e-05,4
0.14955,2.6e+09,-52.855,5.0033e-05,4
0.1507,2.5996e+09,-52.568,5.0067e-05,4
0.15185,2.6002e+09,-52.298,5.0028e-05,4
0.153,2.5999e+09,-52.044,5.0002e-05,4
0.15415,2.5996e+09,-51.865,4.992e-05,4
0.1553,2.5998e+09,-51.591,4.997e-05,4
0.15645,2.5996e+09,-51.336,5.0093e-05,4
0.1576,2.6002e+09,-51.19,5.0095e-05,4
0.15875,2.6005e+09,-50.951,4.9887e-05,4
0.1599,2.5995e+09,-50.693,4.9926e-05,4
0.16105,2.6e+09,-50.461,5.012e-05,4
0.1622,2.6004e+09,-50.302,5.0023e-05,4
0.16335,2.6001e+09,-50.048,4.9884e-05,4
0.1645,2.5997e+09,-49.878,5.0005e-05,4
0.16565,2.5997e+09,-49.791,5.0064e-05,4
0.1668,2.6e+09,-49.524,5.0041e-05,4
0.16795,2.6005e+09,-49.388,5.0036e-05,4
0.1691,2.6004e+09,-49.174,5.0053e-05,4
0.17025,2.5995e+09,-48.996,5.0016e-05,4
0.1714,2.5996e+09,-48.934,4.9977e-05,4
0.17255,2.6e+09,-48.782,5.0048e-05,4
0.1737,2.6005e+09,-48.727,5.0117e-05,4
0.17485,2.5997e+09,-48.641,4.9965e-05,4
0.176,2.6003e+09,-48.484,4.9884e-05,4
0.17715,2.5998e+09,-48.39,4.989e-05,4
0.1783,2.6003e+09,-48.42,4.9925e-05,4
0.17945,2.6005e+09,-48.393,4.9896e-05,4
0.1806,2.5997e+09,-48.316,4.9955e-05,4
0.18175,2.6e+09,-48.215,4.99e-05,4
0.1829,2.5997e+09,-48.125,5.0049e-05,4
0.18405,2.5997e+09,-48.28,4.9908e-05,4
0.1852,2.6004e+09,-48.275,5.0074e-05,4
0.18635,2.6001e+09,-48.135,5.0007e-05,4
0.1875,2.6004e+09,-48.314,4.9951e-05,4
0.18865,2.6e+09,-48.185,4.9916e-05,4
0.1898,2.5995e+09,-48.271,5.0097e-05,4
0.19095,2.5996e+09,-48.395,4.9928e-05,4
0.1921,2.6004e+09,-48.396,4.9988e-05,4
0.19325,2.5997e+09,-48.455,5.0102e-05,4
0.1944,2.6e+09,-48.575,5.0096e-05,4
0.19555,2.5998e+09,-48.663,5.0079e-05,4
0.25,2.6037e+09,-75.123,2.0454e-06,5
0.25075,2.6042e+09,-73.315,1.999e-06,5
0.2515,2.6044e+09,-71.653,1.9823e-06,5
0.25225,2.6038e+09,-69.994,2.0639e-06,5
0.253,2.604e+09,-68.391,1.9963e-06,5
0.25375,2.6043e+09,-67.257,1.9299e-06,5
0.2545,2.6039e+09,-66.164,2.0163e-06,5
0.25525,2.6036e+09,-65.639,1.9588e-06,5
0.256,2.6038e+09,-65.321,1.913e-06,5
0.25675,2.6045e+09,-65.28,2.0615e-06,5
0.2575,2.6039e+09,-65.538,1.9969e-06,5
0.25825,2.6041e+09,-66.246,1.9181e-06,5
0.259,2.6041e+09,-67.172,2.0295e-06,5
0.25975,2.6041e+09,-68.554,2.0292e-06,5
0.2605,2.6043e+09,-69.869,2.0366e-06,5
0.26125,2.6043e+09,-71.51,1.9323e-06,5
0.262,2.6036e+09,-73.437,1.9219e-06,5];

sz = size(dataset);
dataset = dataset(randperm(sz(1)),:);
traindata=dataset(:,1:4);
trainlabel=categorical(dataset(:,5));

classes = unique(trainlabel);
numClasses = numel(unique(trainlabel));

PD = 0.80 ;  
Ptrain = []; Ttrain = [];
Ptest = []; Ttest = [];
for i = 1 : length(classes)
    indi = find(trainlabel==classes(i));
    indi = indi(randperm(length(indi)));
    indj = round(length(indi)*PD);
    Ptrain = [Ptrain; traindata(indi(1:indj),:)]; Ttrain = [Ttrain; trainlabel(indi(1:indj),:)]; 
    Ptest = [Ptest; traindata(indi(1+indj:end),:)]; Ttest = [Ttest; trainlabel(indi(1+indj:end),:)]; 
end
Ptrain2 = []; Ttrain2 = [];
Ptest2 = []; Ttest2 = [];
Ttrain2=categorical(Ttrain);
Ttest2=categorical(Ttest);
for i = 1 : size(Ptrain,1)
    Ptrain2{i} = Ptrain(i,:)';
    %Ttrain2{i} = Ttrain(i);
end
for i = 1 : size(Ptest,1)
    Ptest2{i} = Ptest(i,:)';
    %Ttest2{i} = Ttest(i);
end

layers = [sequenceInputLayer(4)
    convolution1dLayer(3,100,Padding="causal")
    reluLayer
    globalMaxPooling1dLayer
    dropoutLayer
    fullyConnectedLayer(50)
    dropoutLayer
    fullyConnectedLayer(5)
    softmaxLayer
    classificationLayer];

options = trainingOptions('adam', ...
    'MaxEpochs',1000, ...
    'Shuffle','every-epoch', ...
    'Verbose',false, ...
    'ValidationData',{Ptest2,Ttest2},...
    'ExecutionEnvironment', 'auto', ...
    'ValidationPatience',Inf);
[net,dinfo] = trainNetwork(Ptrain2,Ttrain2,layers,options);

figure; plot(dinfo.TrainingAccuracy);

[Med Future]

@yanqi liu what is this picture?

[Med Future]

@yanqi liu this is not training curve

[yanqi liu]

yes，sir，here on web，we can not see the plot curve，so we get the train status info and plot it

this picture is train acc curve by stats info structure