Re: inappropriate upfront?
- From: "Dmitry A. Kazakov" <mailbox@xxxxxxxxxxxxxxxxx>
- Date: Thu, 20 Apr 2006 18:37:25 +0200
On Wed, 19 Apr 2006 19:36:58 +0200, Sasa wrote:
Let us assume that driven by tests, at some point I conclude that I need
some way of specifying decision rules in a declarative manner and build
the engine which applies the rules given the current state of the game.
Obviously there is a number of ways to achieve this. I can develop my
own minimalistic engine, I can use 3rd party library, I could consider
some other language such as Lisp or Prolog or try the interpreter
pattern with OO language (any other ideas?).
You don't necessarily need an interpreter and a brute-force search a-la
Prolog. It really depends on the rules. Rules, like ones of fuzzy control
language (FCL) can be compiled into some internal format. That could be a
decision tree, a NN etc. It could also be other knowledge representation.
You can even use the rules as an input for some supervised learning, train
on the rules as if they were training examples, and use the result as one
rule. I have used this approach in my system.
--
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de
.
- References:
- inappropriate upfront?
- From: Sasa
- Re: inappropriate upfront?
- From: Sasa
- inappropriate upfront?
- Prev by Date: Re: Question on Effective Java Item 27
- Next by Date: Re: UML inner/nested class associations
- Previous by thread: Re: inappropriate upfront?
- Next by thread: sizeof in java
- Index(es):
Relevant Pages
|
