Page 1 of 4

10422 - Knights in FEN

Posted: Fri Mar 21, 2003 2:27 am
by ditrix
What is the best algorithm for this problem? I tried a backtracking and it ran for 2 seconds, but there are more rapid solutions, say instant...

Posted: Fri Mar 21, 2003 4:51 pm
by saiqbal
i used backtracking a with a little bound. i just checked the last move
not to be repeated.my soln. is not fast enough though! :wink:
but i think the best approach is to use BFS. some ppl used IDS also.

thanx
-sohel

Posted: Wed Jun 11, 2003 3:47 am
by Observer
Hello.

I've just solved the problem by DFS. However, my program runs pretty slowly (>> 7 sec)!!!! >_<

I know that there are some much faster methods. Could anyone plz explain them to me??? Many thanks! :roll:


Note: after some simple modification, my program gets ACC in <<1 sec !
However, I'm still looking for faster algorithm... :lol:

Posted: Tue Jun 17, 2003 3:33 am
by Observer
Say, how should I solve this problem within reasonable time if n <= 20??? :confused:

Posted: Tue Jun 17, 2003 9:02 am
by Adrian Kuegel
if N<=20, try bfs starting from starting and final position and check when they meet at some position.

Posted: Tue Jun 17, 2003 10:37 am
by Observer
Could you explain more clearly pls??

By the way, it would be great if the path can be obtained as well! :P

bidirectional bfs

Posted: Tue Jun 17, 2003 2:02 pm
by CelebiX
i used IDA* and solved in 0.000sec.

if n<=20 try bfs from both the start and goal
the bfs tree grown from the start will meet the one grown from the goal at some node x
then start ~> x ~> goal is the path

Re: bidirectional bfs

Posted: Tue Jun 17, 2003 2:14 pm
by Observer
CelebiX wrote:i used IDA* and solved in 0.000sec.
Wow!!!!!!! Really great!!!!!!!!! :lol: :lol: :lol:
CelebiX wrote:if n<=20 try bfs from both the start and goal
the bfs tree grown from the start will meet the one grown from the goal at some node x
then start ~> x ~> goal is the path
This method sounds feasible! But I'm still not so familiar with BFS......
But why would this method be efficient? Could you explain plz?

10422

Posted: Fri Jul 11, 2003 11:14 am
by route
Would anyone tell me how to solve the problem ?

Posted: Fri Jul 11, 2003 12:40 pm
by angga888
Hi route

You can use backtracking to solve this problem. :wink:

angga888

Posted: Sat Jul 12, 2003 4:16 pm
by route
Can you tell me some more details ?
Since there is a restriction that it can only move by exchanging the position with the blank square .

Posted: Mon Aug 04, 2003 4:48 pm
by Observer
Hello! It's me again.

While doing BFS we need to keep a boolean array to see whether a "node" is visited, is that correct?

But in this question, possible configurations = 25!/(12! 12!) = 67603900
So keeping such a huge array is definitely impossible!! (MLE for sure!!!)

So what should I do? Plz help!!!! :cry:

Posted: Tue Aug 05, 2003 8:06 pm
by Whinii F.
I came to see this post eventually and did not solve this problem.

Anyway.. as I understand the post, the size of the state space is pretty large, but by using the method posted above, the number of nodes that you actually have to traverse can become VERY small. :) So you can apply some hashing or clever data structures to check if you have visited a node already or not.

It's a late night in Korea and I'm going to bed now, but when I get an AC I'll come back :)

Posted: Wed Aug 06, 2003 1:40 am
by Observer
I see that the number of "nodes" visited will never exceed 60000......

But how to implement??

Posted: Wed Aug 06, 2003 3:15 am
by Larry
I hashed/precalc from the end position to 11 moves away, and while it's not one of the 0.0 solutions, it's at ~ 0.57 seconds.. so ah well.. good enough for me.. =)