Subject:
|
Re: C$: Knowledge-Based Approach
|
Newsgroups:
|
lugnet.org.ca.rtltoronto
|
Date:
|
Wed, 22 Oct 2003 00:29:41 GMT
|
Viewed:
|
400 times
|
| |
|
|
In lugnet.org.ca.rtltoronto, Steve Hassenplug wrote:
> I've been reading Victor Allis' Knowledge-Based Approach to Connect-4, and I've
> come to the conclusion that is not possible to implement it on an RCX.
Me too. It's definitely brilliant, but not practical to try to teach a computer.
Some parts of his analysis (like the notion of groups and threats, even/odd
rows) are excellent, and not that difficult to combine with a ply-based system.
But most of the rest is just not worth the time or processor power to implement.
The true challenge in the whole thing is working out a really fast algorithm for
evaluating the worth of a given board. No matter whether its loops or structs or
classes or whatever, either way you still have to analyse hundreds of boards
between each turn. So the more you shave down that part of the code, the faster
the whole thing will be. That's my feeling anyways.
btw- I love the moon buggy too... it was awesome to see in person.
|
|
Message is in Reply To:
 | | C$: Knowledge-Based Approach
|
| I've been reading Victor Allis' Knowledge-Based Approach to Connect-4, and I've come to the conclusion that is not possible to implement it on an RCX. Evaluation of his "9 rules" will result in one of three possibilities; 1=you win, 0=undecided, (...) (21 years ago, 21-Oct-03, to lugnet.org.ca.rtltoronto)
|
7 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
