I have a CNN architecture for which I want to optimize the hyperparameters such as learning rate, dropout rate and number of epochs.
I am thinking of a combination of k-fold cross validation and Early Stopping for tuning, because I hope to avoid the search for the optimal number of epochs when I use Early Stopping.
But I'm very unsecure if this approach make sense (because I have to separate an additional validation set for EarlyStopping).
So far, I have not found a definitive statement if this makes sense?
Have you already had experience with this approach?