Is information gain only calculated using decrease in entropy of a dataset?
If I calculate it using the decrease in Gini Impurity, before and after the split, can it not be considered information gain?
Is information gain only calculated using decrease in entropy of a dataset?
If I calculate it using the decrease in Gini Impurity, before and after the split, can it not be considered information gain?
Information gain is a very specific measure in information theory, thus the entropy definition... for sure, the decrease in Gini index can be seen as a proxy for it