View Single Post
Old 08-05-2013, 10:07 AM   #28
Swedish Extrovert
President of the Universe
 
Swedish Extrovert's Avatar
 
Formerly known as mightysmurf

Join Date: Apr 2002
Location: Highlands Ranch
Posts: 15,964

Adopt-a-Bronco:
David Bruton
Default

Quote:
Originally Posted by SeedReaver View Post
You must be a PhD if you know about optimization techniques (though Newton's method is quite simple).

Here is why you need a guess, MightySmurf. Iterative optimization techniques often use gradient based techniques that "find the best direction" to move in to optimize some function. Look at the following picture:



Say you are trying to minimize that wavey function. Many of the techniques Excel uses tries to "greedily" minimize, i.e., find the closest "hill" and move down it. If your initial guess is at point "A," many techniques will stop at the bottom of the valley (where the pink ball is). If you're initial guess is at point "C," you will most likely end up in the valley between "B" and "C." Of course, if you are trying to minimize that function, the valley where the pink ball is located is better than the valley between "B" and "C."

I know nothing about the IRR function or whatever, but
This is really helpful.
Swedish Extrovert is offline   Reply With Quote