Closed hualishiwo closed 4 months ago
This is indeed currently not possible. I have attempted on implementing this for the standard regressors, but this caused a severe performance penalty at prediction. For the sampling approach this should be relatively easy to implement however.
For now I close this issue, if there is a lot of interest in making this work I can take another look.
for extra trees regressor, do you support fit(x, y) while y is of size (n_sample, n_outputs) and n_outputs is not 1? I see in your source code, check_X_y uses multi_output=False, does it mean the multi_output is not supported yet?