3 ms·
What do you do when your (out of sample) test set is from a difference source from the training set, even though your training loss is low?
by zump 10y ago
What do you do when your (out of sample) test set is from a difference source from the training set, even though your training loss is low?
- hatmatrix 10y agoPray. Or build a new training set including a diverse samples so that you are less likely to enter the regime of extrapolation.