Minimum error rate Bayesian decision

Original link: http://blog.csdn.net/angel_yuaner/article/details/47042817

In the general pattern recognition problem, people's goal is often to minimize the classification error, the pursuit of the smallest error rate. According to the previous article, the solution of a decision rule makes: MINP (e) =∫p (e|x) P (x) dx
This is the minimum error rate Bayesian decision.

In the above formula, P (e|x) ≥0,p (x) ≥0 for all x, so MINP (e) is equivalent to all x minimization P (e|x), that is: to maximize the posterior probability P (wi|x). According to the Bayesian formula:
P (wi|x) =p (x|w