Open pedroilidio opened 2 days ago
Current classification criteria for bipartite trees always consider multioutput impurity (average of column-wise impurities). It should not be too complicated to implement a GSO version of them (as if Y were vectorized and columns did not matter).
Current classification criteria for bipartite trees always consider multioutput impurity (average of column-wise impurities). It should not be too complicated to implement a GSO version of them (as if Y were vectorized and columns did not matter).