BFS and DFS were implemented based on the uniformed cost search algorithm; referenced pseudocode from the textbook,"Artificial Intelligence A Modern Approach Third Edition by Stuart J. Russell and Peter Norvig", located in pages.82.
This class implements a backward chaining inference procedure. The implementation is very skeletal, and the resulting reasoning process is not particularly efficient. Knowledge is restricted to the form of definite clauses, grouped into a list of positive literals (facts) and a list of Horn clause implications (rules). The inference procedure maintains a list of goals. On each step, a proof is sought for the first goal in this list, starting by an attempt to unify the goal with any known fact in the knowledge base. If this fails, the rules are examined in the order in which they appear in the knowledge base, searching for a consequent that unifies with the goal. Upon successful unification, a proof is sought for the conjunction of the rule antecedents. If this fails, further rules are considered. Note that this is a strictly depth-first approach, so it is incomplete. Note, also, that there is no backtracking with regard to matches to facts -- the first match is always taken and other potential matches are ignored. This can make the algorithm incomplete in another way. In short, the order in which facts and rules appear in the knowledge base can have a large influence on the behavior of this inference procedure.
In order to use this inference engine, the knowledge base must be initialized by a call to "initKB". Queries are then submitted using the "ask" method. The "ask" function returns a binding list which includes bindings for intermediate variables.
David Noelle -- Tue Apr 10 17:08:45 PDT 2007
This class implements tools for evaluating game states for the Zombie Dice game. These tools include a static function for calculating the expected utility value of a game state using look-ahead to a fixed depth and a static heuristic evaluation function for estimating game payoff values for non-terminal states. Since these tools are all static functions, no objects of this class need to be allocated in order to use them. In general, both heuristic evaluation function values and expected utility values should be between plus and minus "State.win_payoff".
Zombie Dice is a trademark of Steve Jackson Games. For more information about this game, see "zombiedice.sjgames.com".
David Noelle -- Sat Nov 3 16:37:10 PDT 2012
This class implements a layer of processing units. Methods were implemented based on the BACK-PROP-LEARNING neural network algorithm referenced from the textbook, "Artificial Intelligence A Modern Approach Third Edition by Stuart J. Russell and Peter Norvig", located on pages.734.
David Noelle -- Tue Apr 24 15:51:19 PDT 2007
Author of implmentations: Jacky Z. Chen