Train a neural Network with a Table input

Hi i want to train a neural network with a Table in put. The Table is structured as input(rawdata = 111x1 Cell, with each cell a 512x1 Cell) and Response(label = 111x1 Cell). I tried a lot, but it seems to be more compllicated then i thought. I will paste the code Itried so far, but I dont know what to try next.

T = table(rawdata,label);
tblraw = table(rawdata);
tbllabel = (label);
head(T); 
numObservations = size(T, 1);
numObservationsTrain = floor(0.85*numObservations);
numObservationsTest = numObservations - numObservationsTrain;
idx = randperm(numObservations);

idxTrain = idx(1:numObservationsTrain);
idxTest = idx(numObservationsTrain+1:end);
tblTrain = tabulate(idxTrain,:);
tblTest = tabulate(idxTest,:);

This is from a matlab example.

This is the Table

and this is the input in the cell, with different data but same formats:

Thanks, I am noob btw

Answer 1

[Abhaya]

Hi @Mauritz Wilshusen,

To use data from a table containing cell arrays in a neural network, you need to convert the data into a matrix format.

Please refer to the steps given below to prepare the data and train a neural network using the table data.

• Data Preparation: Convert the 'rawdata' column, which is currently stored as a cell array, into a numeric matrix. Each 512x1 cell should be converted into a row, creating a 111x512 matrix.Convert the labels to categorical format.

    X = cell2mat(rawdata);  
    Y = categorical(label); 

• Create neural network architecture: You can design the neural network architecture using the deep learning layers provided by MATLAB. Here is an example of a simple architecture:

    layers = [
        featureInputLayer(512)  % Input layer with 512 features
        fullyConnectedLayer(64) % Hidden layer with 64 neurons
        reluLayer               % Activation function
        fullyConnectedLayer(10) % Output layer with 10 classes (adjust if needed)
        softmaxLayer            % For multi-class classification
        classificationLayer     % Loss layer for classification
        ];

• Set Training Options: Configure the training options using the MATLAB ‘trainingOptions’ function:

    options = trainingOptions('adam', ...
        'MaxEpochs', 50, ...      % Number of epochs
        'MiniBatchSize', 32, ...  % Batch size
        'Shuffle', 'every-epoch', ...
        'Plots', 'training-progress');

• Train the network: You can use the MATLAB ‘trainNetwork’ function to train the network.

    net = trainNetwork(X, Y, layers, options);

For more detailed information, please refer to the MATLAB documentations on ‘trainNetwork’ function and ‘trainingOptions’ function.

• https://www.mathworks.com/help/releases/R2021b/deeplearning/ref/trainnetwork.html
• https://www.mathworks.com/help/releases/R2021b/deeplearning/ref/trainingoptions.html[SS1]

I hope this helps.