A question about harmony measure

[lqm112]

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

[zqf30]

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.