Training data: This data set 'teaches' the algorithm which combinations of pixel values will produce a good predictive model for our distinct classes.
Validation data: This data set, on the other hand, is used to validate the outcome of the classified product by offering an unbiased evaluation of the fitted model.