is there support for segmented or nested models?
era127 opened this issue · 1 comments
I am reading through the MLJ documentation and I don't see an example of how a learning network would implement a segmented or nested model, where an array of models would be fit on segments of the training data.
Similar to a h20 segmented model or R nestedmodels. I also think its possible to do this in mlr3.
Does this functionality exist in MLJ?
Thanks
I think it would be good to add an example along those lines. Have opened this.
A Wolpert stack trains an array of models on different parts of the training data (to splice together "out-of-sample" predictions), which are then forwarded to an adjudicating model for training. So, for an example, you might take a look at this tutorial on rolling your own Stack
(or, probably less accessible, is MLJ's actual implementation of Stack
).