View Single Post
  #2  
Old 04-18-2012, 04:39 PM
htlin's Avatar
htlin htlin is offline
NTU
 
Join Date: Aug 2009
Location: Taipei, Taiwan
Posts: 610
Default Re: Hw 2, problem 3

Quote:
Originally Posted by useral View Post
Hello,

I must be missing the forest for the trees, but I was thinking that the answer should be the weighted sum of the probabilities of (y = f(x)) and (y =/= f(x)), and the weight mu (corresponding to an error, i.e. y =/= f(x)) should be assigned to the probability of that error, i.e. to 1-lambda.

This reasoning gives the answer [d]. Is it flawed?
The cases that h(\mathbf{x}) would be different from y is when either, but not both, of these events happen: h(\mathbf{x}) \neq f(\mathbf{x}), or f(\mathbf{x}) \neq y. Hope this helps.
__________________
When one teaches, two learn.
Reply With Quote