15.08.2013 Views

General Computer Science 320201 GenCS I & II Lecture ... - Kwarc

General Computer Science 320201 GenCS I & II Lecture ... - Kwarc

General Computer Science 320201 GenCS I & II Lecture ... - Kwarc

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 355 (Resolution Proof) We call a resolution refutation P : CNF (A) ⊢R T<br />

a resolution sproof for A ∈ wff o(Vo).<br />

c○: Michael Kohlhase 210<br />

Note: Note that the C-terms in the definition of the resolution calculus are necessary, since<br />

we assumed that the assumptions of the inference rule must match full formulae. The C-terms<br />

are used with the convention that they are optional. So that we can also simplify (A ∨ B) T to<br />

A T ∨ B T .<br />

The background behind this notation is that A and T ∨ A are equivalent for any A. That<br />

allows us to interpret the C-terms in the assumptions as T and thus leave them out.<br />

The resolution calculus as we have formulated it here is quite frugal; we have left out rules for the<br />

connectives ∨, ⇒, and ⇔, relying on the fact that formulae containing these connectives can be<br />

translated into ones without before CNF transformation. The advantage of having a calculus with<br />

few inference rules is that we can prove meta-properties like soundness and completeness with<br />

less effort (these proofs usually require one case per inference rule). On the other hand, adding<br />

specialized inference rules makes proofs shorter and more readable.<br />

Fortunately, there is a way to have your cake and eat it. Derived inference rules have the property<br />

that they are formally redundant, since they do not change the expressive power of the calculus.<br />

Therefore we can leave them out when proving meta-properties, but include them when actually<br />

using the calculus.<br />

Derived Rules of Inference<br />

A1 . . . An<br />

Definition 356 Let C be a calculus, a rule of inference<br />

is called a derived<br />

C<br />

inference rule in C, iff there is a C-proof of A1, . . . , An ⊢ C.<br />

Example 357<br />

Others:<br />

C ∨ (A ⇒ B) T<br />

C ∨ A F ∨ B T<br />

C ∨ (A ⇒ B) T<br />

C ∨ (¬A ∨ B) T<br />

C ∨ ¬A T ∨ B T<br />

C ∨ A F ∨ B T<br />

↦→<br />

C ∨ (A ⇒ B) F<br />

C ∨ A T ; C ∨ B F<br />

C ∨ (A ⇒ B) T<br />

C ∨ A F ∨ B T<br />

C ∨ A ∧ B T<br />

C ∨ A T ; C ∨ B T<br />

c○: Michael Kohlhase 211<br />

C ∨ A ∧ B F<br />

C ∨ A F ∨ B F<br />

With these derived rules, theorem proving becomes quite efficient. To get a better understanding<br />

of the calculus, we look at an example: we prove an axiom of the Hilbert Calculus we have studied<br />

above.<br />

Example: Proving Axiom S<br />

Example 358 Clause Normal Form transformation<br />

113

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

Saved successfully!

Ooh no, something went wrong!