Question on using generative adversarial network GAN for numerical data

Dear All,

Please, i need your support on generative Adversarial Network, i want to use the model on numerical data, but the examples provided by matlab is on image data.

Question 1

As described in the code below, how do i choose the numLatentInputs if i am working on numerical data??

Concerning the projectionSize = [4 4 512]; Please, how do i apply for numerical data

Question 2.

A minibatchqueue function and preprocessing function were introduced has shown the second code below to process and manage mini-batches of data. Since this is applicable for image data, how can i implement minibatches on numerical data?

PS: the data i'm working on has three variables with 100 timestep each.

Thanks.

The link below is the entire code for generative adversarial network (GAN) for images provided by Matlabhttps://www.mathworks.com/help/deeplearning/ug/train-generative-adversarial-network.html

filterSize = 5;
numFilters = 64;
numLatentInputs = 100;

projectionSize = [4 4 512];

layersGenerator = [
    featureInputLayer(numLatentInputs,'Name','in')
    projectAndReshapeLayer(projectionSize,numLatentInputs,'Name','proj');
    
    %% other codes excluded%%.....
    
    augimds.MiniBatchSize = miniBatchSize;

executionEnvironment = "auto";

mbq = minibatchqueue(augimds,...
    'MiniBatchSize',miniBatchSize,...
    'PartialMiniBatch','discard',...
    'MiniBatchFcn', @preprocessMiniBatch,...
    'MiniBatchFormat','SSCB',...
    'OutputEnvironment',executionEnvironment);
    
    %.....other codes excluded.....
    
    function X = preprocessMiniBatch(data)

% Concatenate mini-batch
X = cat(4,data{:});

% Rescale the images in the range [-1 1].
X = rescale(X,-1,1,'InputMin',0,'InputMax',255);

end