Processing math: 100%

Thursday, January 17, 2013

Rice-Shapiro Theorem

The Rice-Shapiro theorem is thus phrased (due to Mr. Zunino's notes, link ), Given \mathcal{F} is a set of partial recursive functions, i.e. \mathcal{F} \subseteq \mathcal{R}. Also A = \{ n | \varphi_n \in \mathcal{F} \} \in \mathcal{RE}. Then f \in \mathcal{F} \Leftrightarrow \exists g \subseteq f . \big( \mathsf{dom}(g) \textit{ is finite} \wedge g \in \mathcal{F} \big)
It is rewritten in the counter-positive manner to prove A \notin \mathcal{RE}.
  • check A is semantically closed
  • (\Rightarrow) pick f \in \mathcal{F} s.t. \nexists g \subseteq f . g \in \mathcal{F} \wedge \mathsf{dom}(g) \textit{ is finite}
  • (\Leftarrow) pick f \notin \mathcal{F} s.t. \exists g \subseteq f . g \in \mathcal{F} \wedge \mathsf{dom}(g) \textit{ is finite}

Some examples.

  • A = \{ n | \mathsf{dom} (\varphi_n) \backslash \mathsf{ran} (\varphi_n) \neq \varnothing \}
  • B = \{ n | \mathsf{dom} (\varphi_n) \text{ is finite} \}
  • C = \{ n | \mathsf{dom} (\varphi_n) \text{ is infinite} \}

A \notin \mathcal{RE}. By R-S \Leftarrow, we provide g(0)=1 \in \mathcal{F}, also we scheme f(0)=1,f(1)=0, s.t. g \subseteq f \wedge f \notin \mathcal{F}, both f and g are finite (let the otherwise case undefined).

B and C are both not \mathcal{RE} (directly from R-S theorem). This says, the set of indices of all finite domain functions are not recursively enumerable. So is the set of indices of all infinite domain ones.

No comments:

Post a Comment