2 ms·
I think true learning requires generalization performance- prediction on outgroups which were not represented in the training set (https://en.wikipedia.org/wiki
by dekhn 7y ago
I think true learning requires generalization performance- prediction on outgroups which were not represented in the training set (https://en.wikipedia.org/wiki/Generalization_error https://en.wikipedia.org/wiki/Generalization_error). It's very similar to when a new theory is able to predict experimental observations that previous theories cannot (general relativity and QM being two great examples).