Invalid training data. Sequence responses must have the same sequence length as the corresponding predictors.

Confused as to why categorical did nto work is there somwhere else that I need to fix

clc; clear all; close all;
load Projectdata.mat

% Split Data Glucose
GlucoseReadings_T = GlucoseReadings';
GlucoseReadings_train = GlucoseReadings_T;
train_GlucoseReadings = GlucoseReadings_train(1:84,:);
train_GR_output = GR_output(1:17);

% Data Batch Glucose
GlucoseReadingsTrain=(reshape(train_GlucoseReadings', [1749,84]));
val_GlucoseReadings = GlucoseReadings_train(85:102,:);
val_GR_output = GR_output(85:102);
GlucoseReadingsVal=(reshape(val_GlucoseReadings', [1749,18]));
test_GlucoseReadings =GlucoseReadings_train(103:120,:);
test_GR_output = GR_output(103:120);
GlucoseReadingsTest=(reshape(test_GlucoseReadings', [1749,18]));
numFeatures = size(GlucoseReadings_T,2);
% number of hidden units represent the size of the data
numHiddenUnits = 120;
%number of classes repsresnt healthy and diabetic
numClasses = length(categories(categorical(GR_output)));
layers = [ ...
    sequenceInputLayer(numFeatures) 
    %dropoutLayer(0.5)
    instanceNormalizationLayer
    bilstmLayer(round(numHiddenUnits/2),'OutputMode','sequence')
    fullyConnectedLayer(numClasses)
    instanceNormalizationLayer
    softmaxLayer
    classificationLayer];
options = trainingOptions('adam', ...
    'MaxEpochs',100, ...
    'GradientThreshold',1, ...
    'Verbose',false, ...
    'ValidationData',{GlucoseReadingsVal, val_GR_output},...
    'LearnRateDropFactor',0.2,...
    'LearnRateDropPeriod',5,...
    'Plots','training-progress');
% Train
whos

  Name                          Size                Bytes  Class                           Attributes

  GR_output                     1x120                7296  string                                    
  GlucoseReadings            1749x120             1679040  double                                    
  GlucoseReadingsTest        1749x18               251856  double                                    
  GlucoseReadingsTrain       1749x84              1175328  double                                    
  GlucoseReadingsVal         1749x18               251856  double                                    
  GlucoseReadings_T           120x1749            1679040  double                                    
  GlucoseReadings_train       120x1749            1679040  double                                    
  INS_output                    1x120                7296  string                                    
  InsulinReadings            1758x120             1687680  double                                    
  layers                        7x1                  3332  nnet.cnn.layer.Layer                      
  numClasses                    1x1                     8  double                                    
  numFeatures                   1x1                     8  double                                    
  numHiddenUnits                1x1                     8  double                                    
  options                       1x1                253569  nnet.cnn.TrainingOptionsADAM              
  test_GR_output                1x18                 1248  string                                    
  test_GlucoseReadings         18x1749             251856  double                                    
  train_GR_output               1x17                 1048  string                                    
  train_GlucoseReadings        84x1749            1175328  double                                    
  val_GR_output                 1x18                 1248  string                                    
  val_GlucoseReadings          18x1749             251856  double                                    

net = trainNetwork(GlucoseReadingsTrain,categorical(train_GR_output),layers,options);

Error using trainNetwork (line 184)
Invalid training data. Sequence responses must have the same sequence length as the corresponding predictors.

% Test
miniBatchSize = 27;
GR_outputPred = classify(net,GlucoseReadingsTest,...
    'MiniBatchSize',miniBatchSize,...
    'Environment','cpu');
acc = mean(GR_outputPred(:) == categorical(test_GR_output(:)))
figure
t = confusionchart(categorical(test_GR_output(:)),GR_outputPred(:));
figure
t1 = confusionchart(categorical(test_INS_output(:)),INS_outputPred(:));

Answer 1

[Walter Roberson]

net = trainNetwork(GlucoseReadingsTrain,categorical(train_GR_output),layers,options);

Your GlucoseReadingsTrain is 1749 x 84

Your train_GR_output is 1 x 18

There isn't anything remotely compatible about those.

You should not be training on a single data set. You should be creating a dataset that has samples from all of your classes, and the corresponding section parameter should indicate which class each sample belongs to.

As I advised you before, you should be creating a variable to hold the number of training samples for each class, and the number of test samples, and you should be creating subsets using code similar to

class1_training = class1_data(1:number_class1_train,:);
class1_test     = class1_data(number_class1_train+1:number_class1_train+number_class1_test,:);
class1_valid    = class1_data(number_class1_train+number_class1_test+1:end,:);
response1_training = class1_response(1:number_class1_train,:);
response1_test     = class1_response(number_class1_train+1:number_class1_train+number_class1_test,:);
response1_valid    = class1_response(number_class1_train+number_class1_test+1:end,:);


%more stuff
training_data = [class1_training; class2_training];
response_data = categorical([response1_training; response2_training]);