Chapter 7 : Gradient Boosting Algorithm : loss function is incorrect

[pschhikara88]

Formula should be :

Incorrect formula
`Li = - yi log(Pi) + (1- yi)log(1-Pi)

Correct formula

Either
Li = -  ( yi log(Pi) + (1- yi)log(1-Pi) )
or

Li = - yi log(Pi) - (1- yi)log(1-Pi)

Because i tried to derive the above incorrect formula to
Li = log(1+ e^yi) - yi*yihat
and then i realised, initial loss function is not correct.

Please let me know if my understanding is correct.

[aditya12343212]

Hey! Just wanted to point out a small but important sign error in the binary cross-entropy loss formula that appears in some parts of the code/docs.

Incorrect version (missing the negative sign on the second term):

$$L_i = -y_i \log(p_i) + (1 - y_i) \log(1 - p_i)$$