1.the information gain of a descriptor reflects the ability of this descriptor to split the data set into two “homogeneous” groups. Homogeneity is determined according to Shannon’s entropy measure.
2.the information gain of a descriptor reflects the amount of information we gain (measure by entropy) of the class labels that can be explained by the descriptor.
Hi Abraham, they both make sense, but here are a couple of suggestions.
I would refer the "degree of purity" of the partitions after splitting rather than homogeneity.
I would specify information gain as measured by a reduction in entropy by splitting.
It really depends who you are writing for - if it is for undergraduates or for a PhD thesis to be examined by domain as well as AI specialists I would keep the wording as simple as possible.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence