When using classifiers and testing the accuracy of the models, how big does the difference between the training set accuracy and testing set accuracy need to be in order to consider the model overfit?
I don't think there is an exact answer for that question. I found some interesting answers here, take a look: https://stats.stackexchange.com/questions/263111/how-much-is-overfitting https://stats.stackexchange.com/questions/294661/train-vs-test-error-gap-and-its-relationship-to-overfitting-reconciling-confli
