I am not sure what do you mean by signals. As per my understanding CNN are used to classify images.
For image classification using Vision Transformer (ViT) networks, you can find detailed instructions and a pretrained ViT network on the MathWorks website here : https://www.mathworks.com/help/vision/ug/transfer-learning-using-pretrained-vit-network.html
Additionally, the MathWorks team has provided an implementation of several variants of the Vision Transformer model on MATLAB Central File Exchange. You can access it here. https://www.mathworks.com/matlabcentral/fileexchange/129739-computer-vision-toolbox-model-for-vision-transformer-network
I recommend exploring the discussion section of the File Exchange entry, where you'll find valuable insights and examples. In particular, there's a code snippet provided by a user named Muhammad that demonstrates how to modify the classification layers of a network.
lgraph = removeLayers(lgraph, {'head','softmax'});
numClass = 100;
newLayers = [
fullyConnectedLayer(numClass,'Name','fc','WeightLearnRateFactor',10,'BiasLearnRateFactor',10)
softmaxLayer('Name','softmax')
classificationLayer('Name','classoutput')];
lgraph = addLayers(lgraph,newLayers);
lgraph = connectLayers(lgraph,'cls_index' ,'fc');
These resources should help you get started with using transformers for classification tasks, whether they involve images or signals / matrix.
