1
$\begingroup$

I am self-studying elements of statistical learning. I got stuck in the following equations: The expected prediction error for classification is given as:

$$G(x) = argmin_{g \in G} \sum _{k=1}^K L(G_k,g) Pr (G_k|X=x) $$ where L is the loss function, G is the set of possible classes, g is our predictions. Then the book says:

with the 0-1 loss function this fun simplifies to:

$$G(x) = argmin_{g \in G}[1-Pr(g|X=x)] $$

I could not understand the simplification.

$\endgroup$
0

1 Answer 1

0
$\begingroup$

If we use the 0/1 loss, the expected error becomes the sum of the probabilities to have a different class than g :

$$G(x) = argmin_{g \in G} \sum _{1 \le k \le K,G_{k}\ne g} Pr (G_k|X=x) $$

Using P(E) = 1 - P(complement of E) :

$$G(x) = argmin_{g \in G}[1-Pr(g|X=x)] $$

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.