Maximum PC

We don't really have a "theory section" and I don't think we should, because the PP isn't a horribly overrun section and it wouldn't hurt to keep them together... so I am going to start a thread on AI. On the homework that I just turned in lastnight, there was this question (With my answer) I am curious as to what everyone's thoughts are:

Q:Sometimes there is no good evaluation function for a problem, but there is a good comparison method: a way to tell whether one node is better than another, without assigning numerical values to either. Show that this is enough to do a best-first search. Is there an analog of A*?

A:The idea behind a best-first search is that it chooses each successor / child node based on some rule that defines which nodes are better than one another, typically a function f(n). As long as the comparison method provides a way to determine if one is better than another then a best-first search can be performed choosing each node in succession based on it being better than another. Due to A* being classified as a best-first search and there is a way to qualify a better node from a less ideal node; one could construct a function/method to track each paths overall status related to how “goodâ€

Not bad. You should probably scratch out a C with that answer. =)

Are you using the green someone and Norvig AI book? If so, which chapter and question number?

I should have some time on Saturday to get into it.

http://aima.cs.berkeley.edu/

I think it was 4.12.

That wasn't the only one on the assignment. I will post my whole HW thing here (should be safe, it was due Wed, and he doesn't give partial credit/late credit...) 4.2, 4.3, 4.11, and the question above, 4.12. His biggest thing is mostly about demonstrating an understanding, after I talked to him... that seemed to be his biggest thing, even if our numbers/answers aren't quite right... although there isn't much in the way of partial credit, so either you get it or you don't.


This algorithm will be found to be optimal for any values of w that make the h(n) portion of the function negligible if it is not already admissible. If it is overestimating the distance from the goal it would be made optimal for w = 0. This happens because w = 0 removes h(n) from the decision function removing the effect of overestimating the distance. This also makes it most like a uniform-cost search, which is optimal by its definition. For w = 1 it will also be optimal provided that h(n) is admissible. If h(n) is admissible it will also be optimal for w = 2 as it is going to choose the nodes who are closest to the goal and the factor of two out in front will not change the order they are expanded in.

For w = 0 :: f(n)=2*g(n)
This removes the “estimated distance from the goalâ€

I am dying to get into this... but life keeps throwing up road blocks... :\

Ditto.

I love having great conversations in this folder, and I bitch about their absence when I have the time to participate. Then, when they pop up, I find I don't have time to contribute.

Sorry Crash ...

Don't worry, this class has given me plenty of ammo. Fortunately, I figured out his trick, or at least a trick, to get a decent grade on the HW.

He is kind of stingy.

I have my A* in LISP written...