Bayesian decision theory

Under continuous feature x we have observed,
assume
${\omega_1, \omega_2, ...\omega_c}$ ... $c$ finite categories
${\alpha_1, \alpha_2, ...\alpha_a}$ ... $a$ finite possible actions
then
loss function $\lambda(\alpha_i \vert \omega_j)$ ... the loss caused by taking an action
$\alpha_i$ when $|omega=j$ is the situation.


see paper

-> likelihood ratio

two-category classification

=> discriminant function