Number of elements must not change. Use [] as one of the size inputs to automatically calculate the appropriate size for that dimension.

clc; clear all; close all;
%Import/Upload data
load GlucoseReadings.mat
% change to label vector
CS = categories(categorical(GR_output));
Z1 = []; Z2 = [];
for i = 1 : length(GR_output)
    Z1(i,1) = find(GR_output(i)==CS);
end
%for i = 1 : length(Y2)
    %Z2(i,1) = find(Y2(i)==CS);
%end
Yo1 = GR_output;
%Yo2 = Y2;
GR_output= Z1;
%Y2 = Z2;
%transposing glucose data
GlucoseReadings_T = GlucoseReadings';
%Shuffling data to take randomly
rand('seed', 0)
ind = randperm(size(GlucoseReadings_T, 1));
GlucoseReadings_T = GlucoseReadings_T(ind, :);
GR_output = GR_output(ind);
%Separating data in training, validation and testing data
GlucoseReadings_train = GlucoseReadings_T;
%Partioning data for training 70%
train_GlucoseReadings = GlucoseReadings_train(1:17,:);
%Corresponding X(input) data to Y(output) data
train_GR_output = GR_output(1:17);
%reshaping data into 4D array
GlucoseReadingsTrain=(reshape(train_GlucoseReadings', [1438,1,1,17])); 
%Separating and partioning for validation data 15%
val_GlucoseReadings = GlucoseReadings_train(18:21,:);
%Corresponding X(input) data to Y(output) data
val_GR_output = GR_output(18:21);
%reshaping data into 4D array
GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1438,1,1,18])); %Train data
%Separating and partioning for test data 15%
test_GlucoseReadings = GlucoseReadings_train(19:24,:);
%Corresponding X(input) data to Y(output) data
test_GR_output = GR_output(19:24);
%reshaping data into 4D array
GlucoseReadingsTest=(reshape(GlucoseReadings_X1', [1438,1,1,18])); %Train data
%% NETWORK ARCHITECTURE
layers = [imageInputLayer([1438 1 1])  % Creating the image layer
 convolution2dLayer([102 1],3,'Stride',1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
    dropoutLayer
    fullyConnectedLayer(1)
    regressionLayer];
% Specify training options.
opts = trainingOptions('sgdm', ...
    'MaxEpochs',1500, ...
    'Shuffle','every-epoch', ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'ValidationData',{GlucoseReadingsVal,val_GR_output},...
    'LearnRateDropFactor',0.2,...
    'LearnRateDropPeriod',5,...
    'ExecutionEnvironment', 'cpu', ...
    'ValidationPatience',Inf);
%% Train network
%net = trainNetwork(XTrain,Trainoutfinal,layers,opts);
yc = train_GR_output(:);
net1 = trainNetwork(GlucoseReadingsTrain,yc,layers,opts);
%% Compare against testing Data


Ypredicted = predict(net1, GlucoseReadingsTest)
predictionError = test_GR_output - GR_outputpredicted;
squares = predictionError.^2;
rmse = sqrt(mean(squares))

figure
scatter(GR_outputpredicted, test_GR_output,'+')
title ('True value vs Predicted Value')
xlabel ("Predicted Value")
ylabel ("True Value")

hold on
plot([-3 3], [-7 7], 'b--')

[Nathaniel Porter]

Data used

[KSSV]

GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1438,1,1,18])); %Train data

The above line is not needed. You are trying to put some data aside for validation. The dimensions of this should be same as GlucoseReadingsTrain. If you have other data assign here, if not put some data aside from training data.

This line:

GlucoseReadingsTest=(reshape(GlucoseReadings_X1', [1438,1,1,18])); %Train data

From where the variable GlucoseReadings_X1 has come?

[Nathaniel Porter]

I think that was an error

[Nathaniel Porter]

So do you mean it should be [1438, 1, 1, 17] also?

Answer 1

[Walter Roberson]

fileurl = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/902755/GlucoseReadings.mat';
%Import/Upload data
tname = tempname() + ".mat";
urlwrite(fileurl, tname);
data = load(tname);
GR_output = data.GR_output;
GlucoseReadings = data.GlucoseReadings;

% change to label vector
CS = categories(categorical(GR_output));
Z1 = []; Z2 = [];
for i = 1 : length(GR_output)
    Z1(i,1) = find(GR_output(i)==CS);
end
%for i = 1 : length(Y2)
    %Z2(i,1) = find(Y2(i)==CS);
%end
Yo1 = GR_output;
%Yo2 = Y2;
GR_output= Z1;
%Y2 = Z2;
%transposing glucose data
GlucoseReadings_T = GlucoseReadings';
%Shuffling data to take randomly
rand('seed', 0)
ind = randperm(size(GlucoseReadings_T, 1));
GlucoseReadings_T = GlucoseReadings_T(ind, :);
GR_output = GR_output(ind);
%Separating data in training, validation and testing data
GlucoseReadings_train = GlucoseReadings_T;
%Partioning data for training 70%
train_GlucoseReadings = GlucoseReadings_train(1:17,:);
%Corresponding X(input) data to Y(output) data
train_GR_output = GR_output(1:17);
%reshaping data into 4D array
GlucoseReadingsTrain=(reshape(train_GlucoseReadings', [1438,1,1,17])); 
%Separating and partioning for validation data 15%
val_GlucoseReadings = GlucoseReadings_train(18:21,:);

That is obviously 4 by something

%Corresponding X(input) data to Y(output) data
val_GR_output = GR_output(18:21);
%reshaping data into 4D array
whos

  Name                          Size                 Bytes  Class     Attributes

  CS                            2x1                    238  cell                
  GR_output                    24x1                    192  double              
  GlucoseReadings            1438x24                276096  double              
  GlucoseReadingsTrain       1438x1x1x17            195568  double              
  GlucoseReadings_T            24x1438              276096  double              
  GlucoseReadings_train        24x1438              276096  double              
  Yo1                           1x26                  1518  string              
  Z1                           26x1                    208  double              
  Z2                            0x0                      0  double              
  ans                           1x47                    94  char                
  data                          1x1                 277950  struct              
  fileurl                       1x89                   178  char                
  i                             1x1                      8  double              
  ind                           1x24                   192  double              
  tname                         1x1                    308  string              
  train_GR_output              17x1                    136  double              
  train_GlucoseReadings        17x1438              195568  double              
  val_GR_output                 4x1                     32  double              
  val_GlucoseReadings           4x1438               46016  double              

and checking we see that Yes, it is 4 x something

GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1438,1,1,18])); %Train data

Error using reshape
Number of elements must not change. Use [] as one of the size inputs to automatically calculate the appropriate size for that dimension.

The code is treating it as if it is 18 x something, not 4 x something.

But why 18 anyhow? You pulled off 17 columns to make GclusoseReadingsTrain, and you pulled off 4 columns to make val_GlucoseReadings, and columns 22 to 24 are not extracted from yet... but why pull out 18 ? It would make more sense if you were doing 18:end to pull out everything other than the first 17.

%Separating and partioning for test data 15%
test_GlucoseReadings = GlucoseReadings_train(19:24,:);

Why are you re-using 19, 20, 21, as well as adding in 22, 23, 24, but at the same time ignoring 18 ??

I would recommend to you that you use variables to hold the number you are selecting for training, and then be consistent about using indices like 1:numtrain, numtrain+1:numtrain+numtest, numtrain+numtest+1:end

%Corresponding X(input) data to Y(output) data
test_GR_output = GR_output(19:24);
%reshaping data into 4D array
GlucoseReadingsTest=(reshape(GlucoseReadings_X1', [1438,1,1,18])); %Train data
%% NETWORK ARCHITECTURE
layers = [imageInputLayer([1438 1 1])  % Creating the image layer
 convolution2dLayer([102 1],3,'Stride',1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
    dropoutLayer
    fullyConnectedLayer(1)
    regressionLayer];
% Specify training options.
opts = trainingOptions('sgdm', ...
    'MaxEpochs',1500, ...
    'Shuffle','every-epoch', ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'ValidationData',{GlucoseReadingsVal,val_GR_output},...
    'LearnRateDropFactor',0.2,...
    'LearnRateDropPeriod',5,...
    'ExecutionEnvironment', 'cpu', ...
    'ValidationPatience',Inf);
%% Train network
%net = trainNetwork(XTrain,Trainoutfinal,layers,opts);
yc = train_GR_output(:);
net1 = trainNetwork(GlucoseReadingsTrain,yc,layers,opts);
%% Compare against testing Data


Ypredicted = predict(net1, GlucoseReadingsTest)
predictionError = test_GR_output - GR_outputpredicted;
squares = predictionError.^2;
rmse = sqrt(mean(squares))

figure
scatter(GR_outputpredicted, test_GR_output,'+')
title ('True value vs Predicted Value')
xlabel ("Predicted Value")
ylabel ("True Value")

hold on
plot([-3 3], [-7 7], 'b--')

[Nathaniel Porter]

clc; clear all; close all;
%Import/Upload data
load GlucoseReadings.mat

Error using load
Unable to find file or directory 'GlucoseReadings.mat'.

% change to label vector
CS = categories(categorical(GR_output));
Z1 = []; Z2 = [];
for i = 1 : length(GR_output)
    Z1(i,1) = find(GR_output(i)==CS);
end
%for i = 1 : length(Y2)
    %Z2(i,1) = find(Y2(i)==CS);
%end
Yo1 = GR_output;
%Yo2 = Y2;
GR_output= Z1;
%Y2 = Z2;
%transposing glucose data
GlucoseReadings_T = GlucoseReadings';
%Shuffling data to take randomly
rand('seed', 0)
ind = randperm(size(GlucoseReadings_T, 1));
GlucoseReadings_T = GlucoseReadings_T(ind, :);
GR_output = GR_output(ind);
%Separating data in training, validation and testing data
GlucoseReadings_train = GlucoseReadings_T;
%Partioning data for training 70%
train_GlucoseReadings = GlucoseReadings_train(1:17,:);
%Corresponding X(input) data to Y(output) data
train_GR_output = GR_output(1:17);
%reshaping data into 4D array
GlucoseReadingsTrain=(reshape(train_GlucoseReadings', [1438,1,1,17])); 
%Separating and partioning for validation data 15%
val_GlucoseReadings = GlucoseReadings_train(18:21,:);
%Corresponding X(input) data to Y(output) data
val_GR_output = GR_output(18:21);
%reshaping data into 4D array
GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1438,1,1,4])); %Train data
%Separating and partioning for test data 15%
test_GlucoseReadings = GlucoseReadings_train(19:24,:);
%Corresponding X(input) data to Y(output) data
test_GR_output = GR_output(19:24);
%reshaping data into 4D array
GlucoseReadingsTest=(reshape(test_GlucoseReadings', [1438,1,1,4])); %Train data
%% NETWORK ARCHITECTURE
layers = [imageInputLayer([1438 1 1])  % Creating the image layer
 convolution2dLayer([102 1],3,'Stride',1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
    dropoutLayer
    fullyConnectedLayer(1)
    regressionLayer];
% Specify training options.
opts = trainingOptions('sgdm', ...
    'MaxEpochs',1500, ...
    'Shuffle','every-epoch', ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'ValidationData',{GlucoseReadingsVal,val_GR_output},...
    'LearnRateDropFactor',0.2,...
    'LearnRateDropPeriod',5,...
    'ExecutionEnvironment', 'cpu', ...
    'ValidationPatience',Inf);
%% Train network
%net = trainNetwork(XTrain,Trainoutfinal,layers,opts);
yc = train_GR_output(:);
net1 = trainNetwork(GlucoseReadingsTrain,yc,layers,opts);
%% Compare against testing Data


Ypredicted = predict(net1, GlucoseReadingsTest)
predictionError = test_GR_output - GR_outputpredicted;
squares = predictionError.^2;
rmse = sqrt(mean(squares))

figure
scatter(GR_outputpredicted, test_GR_output,'+')
title ('True value vs Predicted Value')
xlabel ("Predicted Value")
ylabel ("True Value")

hold on
plot([-3 3], [-7 7], 'b--')

Is this what you meant I should fix by chance

[Walter Roberson]

You are still making mistakes in selecting your training, validation, and test set.

Create a variable which is the number of training, and index 1:numtrain instead of 1:17.

Create a variable which is the number of validations, and index numtrain+1:numtrain+numval instead of 18:21

Create a variable which is the number of test, as size(GlocuseReadings_T,1) - numtrain - numval, and index with numtrain+numval+1:numtrain+numval+numtest instead of 19:24

And in code like [1438,1,1,17] replace the 17 with numtrain, and replace the 4's with numval

fileurl = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/902755/GlucoseReadings.mat';
%Import/Upload data
tname = tempname() + ".mat";
urlwrite(fileurl, tname);
data = load(tname);
GR_output = data.GR_output;
GlucoseReadings = data.GlucoseReadings;

% change to label vector
CS = categories(categorical(GR_output));
Z1 = []; Z2 = [];
for i = 1 : length(GR_output)
    Z1(i,1) = find(GR_output(i)==CS);
end
%for i = 1 : length(Y2)
    %Z2(i,1) = find(Y2(i)==CS);
%end
Yo1 = GR_output;
%Yo2 = Y2;
GR_output= Z1;
%Y2 = Z2;
%transposing glucose data
GlucoseReadings_T = GlucoseReadings';
%Shuffling data to take randomly
rand('seed', 0)
ind = randperm(size(GlucoseReadings_T, 1));
GlucoseReadings_T = GlucoseReadings_T(ind, :);
GR_output = GR_output(ind);
%Separating data in training, validation and testing data
GlucoseReadings_train = GlucoseReadings_T;
%Partioning data for training 70%
train_GlucoseReadings = GlucoseReadings_train(1:17,:);
%Corresponding X(input) data to Y(output) data
train_GR_output = GR_output(1:17);
%reshaping data into 4D array
GlucoseReadingsTrain=(reshape(train_GlucoseReadings', [1438,1,1,17])); 
%Separating and partioning for validation data 15%
val_GlucoseReadings = GlucoseReadings_train(18:21,:);
%Corresponding X(input) data to Y(output) data
val_GR_output = GR_output(18:21);
%reshaping data into 4D array
GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1438,1,1,4])); %Train data
%Separating and partioning for test data 15%
test_GlucoseReadings = GlucoseReadings_train(19:24,:);
%Corresponding X(input) data to Y(output) data
test_GR_output = GR_output(19:24);
%reshaping data into 4D array
whos

  Name                          Size                 Bytes  Class     Attributes

  CS                            2x1                    238  cell                
  GR_output                    24x1                    192  double              
  GlucoseReadings            1438x24                276096  double              
  GlucoseReadingsTrain       1438x1x1x17            195568  double              
  GlucoseReadingsVal         1438x1x1x4              46016  double              
  GlucoseReadings_T            24x1438              276096  double              
  GlucoseReadings_train        24x1438              276096  double              
  Yo1                           1x26                  1518  string              
  Z1                           26x1                    208  double              
  Z2                            0x0                      0  double              
  ans                           1x47                    94  char                
  data                          1x1                 277950  struct              
  fileurl                       1x89                   178  char                
  i                             1x1                      8  double              
  ind                           1x24                   192  double              
  test_GR_output                6x1                     48  double              
  test_GlucoseReadings          6x1438               69024  double              
  tname                         1x1                    308  string              
  train_GR_output              17x1                    136  double              
  train_GlucoseReadings        17x1438              195568  double              
  val_GR_output                 4x1                     32  double              
  val_GlucoseReadings           4x1438               46016  double              

GlucoseReadingsTest=(reshape(test_GlucoseReadings', [1438,1,1,4])); %Train data

Error using reshape
Number of elements must not change. Use [] as one of the size inputs to automatically calculate the appropriate size for that dimension.

%% NETWORK ARCHITECTURE
layers = [imageInputLayer([1438 1 1])  % Creating the image layer
 convolution2dLayer([102 1],3,'Stride',1)
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
    dropoutLayer
    fullyConnectedLayer(1)
    regressionLayer];
% Specify training options.
opts = trainingOptions('sgdm', ...
    'MaxEpochs',1500, ...
    'Shuffle','every-epoch', ...
    'Plots','training-progress', ...
    'Verbose',false, ...
    'ValidationData',{GlucoseReadingsVal,val_GR_output},...
    'LearnRateDropFactor',0.2,...
    'LearnRateDropPeriod',5,...
    'ExecutionEnvironment', 'cpu', ...
    'ValidationPatience',Inf);
%% Train network
%net = trainNetwork(XTrain,Trainoutfinal,layers,opts);
yc = train_GR_output(:);
net1 = trainNetwork(GlucoseReadingsTrain,yc,layers,opts);
%% Compare against testing Data


Ypredicted = predict(net1, GlucoseReadingsTest)
predictionError = test_GR_output - GR_outputpredicted;
squares = predictionError.^2;
rmse = sqrt(mean(squares))

figure
scatter(GR_outputpredicted, test_GR_output,'+')
title ('True value vs Predicted Value')
xlabel ("Predicted Value")
ylabel ("True Value")

hold on
plot([-3 3], [-7 7], 'b--')