Open lqm112 opened 7 months ago
I think that using the former definition can ensure that h is not greater than 1. Compared to the latter, using the former enables the loss function to give relatively more attention to the features that are hard to be classified.
Why is the definition of harmony measure h = p ^ b u ^ ( 1-b ), not h = b p + ( 1-b ) * u,Can you explain it in detail