Error using trainnet (line 46)

Dear sir;

My XTrain 48941x1 cell and TTrain 48941x1 categorical as a shown at below

why does I get this error?

Error using trainnet (line 46)

Number of observations in predictors (48941) and targets (1) must match. Check that

the data and network are consistent.

layers = [
    sequenceInputLayer([30 30 1],'Name','input')    % For 2-D image sequence input, InputSize is vector of three elements [h w c], where h is the image height, w is the image width, and c is the number of channels of the image.
    convolution2dLayer(5,8,'Stride',1,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    convolution2dLayer(5,8,'Stride',2,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    convolution2dLayer(5,8,'Stride',1,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    convolution2dLayer(5,16,'Stride',2,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    convolution2dLayer(5,16,'Stride',1,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    convolution2dLayer(5,32,'Stride',2,'Padding','same','WeightsInitializer','he','Name','conv','DilationFactor',1);
    batchNormalizationLayer('Name','bn')                             % A batch normalization layer normalizes a mini-batch of data across all observations for each channel independently. To speed up training of the convolutional neural network and reduce the sensitivity to network initialization, use batch normalization layers between convolutional layers and nonlinearities, such as ReLU layers. After normalization, the layer scales the input with a learnable scale factor γ and shifts it by a learnable offset β.
    reluLayer('Name','Relu')                                           % An ELU activation layer performs the identity operation on positive inputs and an exponential nonlinearity on negative inputs.
    globalAveragePooling2dLayer(Name="gap1")
    fullyConnectedLayer(7)
    softmaxLayer];

options = trainingOptions("adam", ...
    MaxEpochs=4, ...
    InitialLearnRate=0.002,...
    MiniBatchSize=128,...
    GradientThreshold=1, ...
    LearnRateSchedule="piecewise", ...
    LearnRateDropPeriod=20, ...
    LearnRateDropFactor=0.8, ...
    L2Regularization=1e-3,...
    Shuffle="every-epoch", ...
    Plots="training-progress", ...
    ObjectiveMetricName="loss", ...
    OutputNetwork="best-validation", ...
    ValidationPatience=5, ...              % Specify the validation patience as 5 so training stops if the recall has not decreased for five iterations.
    ValidationFrequency=50, ... 
    Verbose=false, ...
    Metrics="accuracy", ...
    ValidationData={XValidation,TValidation});

net = trainnet(XTrain,TTrain,layers,"crossentropy",options);

[Matt J]

My XTrain 48941x1 cell and TTrain 48941x1 categorical as a shown at below

Please attach them (or non-proprietory data of the same structure) so that we can try to reproduce the error.

[Bahadir]

Dear @Matt J

XTrain (48941x1 cell) file is 355 MB.

Therefore I attached XTrain (700x1 cell) and TTrain(700x1 categorical) with first 700 row.

I get same fault.

Error using trainnet (line 46)

Number of observations in predictors (700) and targets (1) must match. Check that the data and

network are consistent.

[Walter Roberson]

In order to test we would need corresponding XValidation and TValidation

Answer 1

[Matt J]

It appears that if your XTrain is in cell array form, you need to put your TTrain data in cell form as well:

load('attachedData.mat'); clear ans;  whos  %Inventory

  Name           Size              Bytes  Class                           Attributes

  TTrain       200x1                1154  categorical                               
  XTrain       200x1             1464000  cell                                      
  layers        22x1               23222  nnet.cnn.layer.Layer                      
  options        1x1             1442212  nnet.cnn.TrainingOptionsADAM              

TTrain=num2cell(TTrain);

options.Plots='none'; %Online environment doesn't support plots
options.Verbose=true;
options.ValidationData={XTrain,TTrain};  %Fake validation data


testPrediction=minibatchpredict(dlnetwork(layers), XTrain(1:3)) %test

testPrediction = 1×7×3 single array
testPrediction(:,:,1) =

    0.0716    0.1252    0.2103    0.1289    0.1714    0.1264    0.1662


testPrediction(:,:,2) =

    0.0835    0.1331    0.1854    0.1318    0.1799    0.1212    0.1652


testPrediction(:,:,3) =

    0.0935    0.1419    0.1846    0.1289    0.1629    0.1257    0.1625

net = trainnet(XTrain,TTrain,layers,"crossentropy",options);

    Iteration    Epoch    TimeElapsed    LearnRate    TrainingLoss    ValidationLoss    TrainingAccuracy    ValidationAccuracy
    _________    _____    ___________    _________    ____________    ______________    ________________    __________________
            0        0       00:00:02        0.002                            2.0279                                        15
            1        1       00:00:02        0.002          2.0175                                22.656                      
            4        4       00:00:04        0.002          1.6088            1.8219              52.344                  44.5
Training stopped: Max epochs completed

[Matt J]

You are using a sequenceInputLayer, but your training inputs appear to just be 30x30 images. An imageInputLayer might be more appropriate...

load('attachedData.mat'); 

XTrain=cat(4,XTrain{:});
layers(1)=imageInputLayer([30,30,1],Name="input");

options.Plots='none'; %Online environment doesn't support plots
options.Verbose=true;
options.ValidationData={XTrain,TTrain};  %Fake validation data

testPrediction=minibatchpredict(dlnetwork(layers), XTrain(:,:,:,1:3)) %test

testPrediction = 3×7

0.2280 0.2614 0.1365 0.0465 0.0868 0.0722 0.1687 0.2249 0.2505 0.1357 0.0480 0.1050 0.0783 0.1576 0.2297 0.2142 0.1242 0.0550 0.1048 0.0976 0.1744

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

net = trainnet(XTrain,TTrain,layers,"crossentropy",options);

    Iteration    Epoch    TimeElapsed    LearnRate    TrainingLoss    ValidationLoss    TrainingAccuracy    ValidationAccuracy
    _________    _____    ___________    _________    ____________    ______________    ________________    __________________
            0        0       00:00:02        0.002                            2.0273                                        20
            1        1       00:00:02        0.002          1.9511                                28.906                      
            4        4       00:00:03        0.002          1.4855            1.7872              73.438                    28
Training stopped: Max epochs completed

[Bahadir]

Thanks it is OK