Closed MauroLuzzatto closed 3 years ago
yeah, I think this would generally work. In practice it may be trickier though due to the nature of regression values (compared to class label probabilities in range (0, 1)). The implementation may have to be slightly tweaked for regression but I am not 100% sure; haven't tried.
Great, thanks for the fast response. I will look into this!
Thanks a lot, to the creators for the nice implementation of the
create_counterfactual
!I was wondering, if it is reasonable to use the implementation also for regression task? As for example if
y_desired
could also be a float, which should be reached? I tried it out and the results seem reasonable, nevertheless, I quickly want to ask, if this usage would make sense. Thanks a lot!