search

soulmachine / machine-learning-cheat-sheet

Classical equations and diagrams in machine learning
http://soulmachine.me
7.44k stars 1.27k forks source link

Why logarithmic loss function no need to calcalute the distance between Y and f(X)? #10

Open jackeylu opened 9 years ago

jackeylu commented 9 years ago

in chapterIntroduction.tex, the Logarithmic loss function is defined as \ $L(Y,P(Y|X))=-\log{P(Y|X)}$.

I think f(X) = P(Y|X), and why L(Y, f(X)) = -log(f(X)) can be used as a loss function? It can not tell us the loss between the predicate value and real value.

Hope for your reply.