06.03.2013 Views

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Definition 5.10: A proof process is called complete, if for any inference<br />

α, that follows logically from a given set of axioms S, i..e., S α, the proof<br />

procedure can prove α, i.e., Sα.<br />

Theorem 5.1: The resolution theorem is sound.<br />

Proof: Given a set of clauses S <strong>and</strong> a goal α. Suppose we derived α from S<br />

by the resolution theorem. By our usual notation, we thus have S α. We<br />

want to prove that the derivation is logically sound, i.e., S α. Let us prove<br />

the theorem by the method of contradiction. So, we presume that the<br />

consequent S α is false, which in other words means S ¬ α. Thus ¬ α is<br />

satisfiable. To satisfy it, we assign truth values (true / fa se) to all<br />

propositions that are used in α. We now claim that for such assignment,<br />

resolution of any two clauses from S will be true. Thus the resulting clause<br />

even after exhaustion of all clauses through resolution will not be false. Thus<br />

S α is a contradiction. Hence, the assumption S ¬ α is false, <strong>and</strong><br />

consequently S α is true. This is all about the proof [5]. <br />

Theorem 5.2: The resolution theorem is complete.<br />

Proof: Let α be a formula, such that from a given set of clauses S, we have<br />

S α, i.e., α can be logically proved from S. We have to show there exists a<br />

proof procedure for α, i.e., S α.<br />

We shall prove it by the method of contradiction, i.e. let S α not<br />

follow, i.e., S ¬α. In words α is not derivable by a proof procedure from S.<br />

Therefore, S1 S ∪ α is unsatisfiable. We now use an important theorem,<br />

called the ground resolution theorem, that states “if a set of ground clauses<br />

(clauses with no variables) is unsatisfiable, then the resolution closure of those<br />

clauses contains the ‘false’ clause. Thus as S1 is unsatisfiable, the resolution<br />

closure of S1 yields the null clause, which causes a contradiction to S α.<br />

Thus the assumption is wrong <strong>and</strong> hence S α is true. <br />

We now prove the ground resolution theorem, stated below.<br />

Theorem 5.3: If a set of ground clauses S is unsatisfiable, then the<br />

resolution closure T of those clauses contains the false clause.<br />

Proof: We prove the theorem by the method of contradiction. So, we<br />

presume that resolution closure T does not contain false clause <strong>and</strong> will<br />

terminate the proof by showing that S is satisfiable.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!