Matlab "trainNetwork" error Predictors and responses must have the same number of observations

Hi

I am using one of matlab dataset, (transmissionCasingData.csv), to use 1D convolution layer to train a network. Eventhough, the size of my predictors and response are the same, but matlab throws the error that the size of the predictors and response must be the same.

I am wondering if anyone have any idea how to resole the issue.

clear

clc

filename = "transmissionCasingData.csv";

tbl = readtable(filename,'TextType','String');

labelName = "GearToothCondition";

tbl = convertvars(tbl,labelName,'categorical');

categoricalInputNames = ["SensorCondition" "ShaftCondition"];

tbl = convertvars(tbl,categoricalInputNames,'categorical');

for i = 1:numel(categoricalInputNames)

name = categoricalInputNames(i);

oh = onehotencode(tbl(:,name));

tbl = addvars(tbl,oh,'After',name);

tbl(:,name) = [];

end

tbl = splitvars(tbl);

classNames = categories(tbl{:,labelName});

numObservations = size(tbl,1);

numObservationsTrain = floor(0.85*numObservations);

numObservationsTest = numObservations - numObservationsTrain;

idx = randperm(numObservations);

idxTrain = idx(1:numObservationsTrain);

idxTest = idx(numObservationsTrain+1:end);

tblTrain = tbl(idxTrain,:);

tblTest = tbl(idxTest,:);

numFeatures = size(tbl,2) - 1;

numClasses = numel(classNames);

%% Define Layers

classificationLayer];

%}

numFilters = 64;

filterSize = 5;

layers = [

% featureInputLayer(numFeatures)

sequenceInputLayer(numFeatures)

convolution1dLayer(filterSize,numFilters,Padding="causal")

convolution1dLayer(24,3,Padding="causal")

convolution1dLayer(24,3,Padding="causal")

convolution1dLayer(24,3,Padding="causal")

dropoutLayer(0.2)

convolution1dLayer(128,3,Padding="causal")

convolution1dLayer(128,3,Padding="causal")

convolution1dLayer(128,3,Padding="causal")

maxPooling1dLayer(3,Padding="same")

dropoutLayer(0.2)

reluLayer

softmaxLayer

classificationLayer]

miniBatchSize = 16;

options = trainingOptions('adam', ...

'MiniBatchSize',miniBatchSize, ...

'Shuffle','every-epoch', ...

'Plots','training-progress', ...

'Verbose',false);

% Op=table2cell(tblTrain);

TragetData=(tblTrain.GearToothCondition);

% TragetData=table2cell(TragetData);

TrainData=(tblTrain(:,1:22));

TrainData_Cell=(table2cell(TrainData));

% net = trainNetwork(tblTrain,layers,options);

TrainData=(TrainData_Cell');

ResponseData=TragetData';

net = trainNetwork(TrainData,ResponseData,layers,options); % error happnes here

YPred = classify(net,tblTest,'MiniBatchSize',miniBatchSize);

YTest = tblTest{:,labelName};

accuracy = sum(YPred == YTest)/numel(YTest)

%{

The error is:

Error using trainNetwork (line 184)

Invalid training data. Predictors and responses must have the same number of

observations.

Error in test (line 85)

net = trainNetwork(TrainData,ResponseData,layers,options);

%}

[Pratyush Roy]

Hi,

Can you please share the csv file so that I can reproduce the issue on my end?

Thanks.

[SM]

Hi,

The "transmissionCasingData.csv" is an inbuild MATLAB dataset, and you should be able to access it by default.

However, I have attached the '.mat' file in case you do not have access it.

I am using MATLAB 2021b.

Thank you for your help in advance

Answer 1

[Kumar Pallav]

The error is generally caused due to mismatch in shapes in the data provided to the trainNetwork. You may refer to a similar problem here to resolve the issue.