This message was deleted.

Hey. Each model works independently. How would your approach work? Do you compute the predictions from model1 on a holdout set and train model2 on that?

That's right, training model2 on train set + prediction from model1. It's the same idea as in AR, but using N models instead of 1

Got it. There's no built-in way to do that, but you can definitely do it manually by modifying the training set in a loop