Reduction and Degree Reduction and Degree Relative Computability Relative Computability Turing Reducibility Turing Reducibility Outline
∗ Reducibility and Degree 1 Reduction and Degree Many-One Reduction Degrees Xiaofeng Gao m-Complete r.e. Set
Department of Computer Science and Engineering Shanghai Jiao Tong University, P.R.China 2 Relative Computability CS363-Computability Theory 3 Turing Reducibility
∗ Special thanks is given to Prof. Yuxi Fu for sharing his teaching materials. CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 1/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 2/64
Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set General Remark Many-One Reduction
A problem is a set of numbers.
A reduction is a way of defining a solution of a problem with the help The set A is many-one reducible (m-reducible) to the set B if there is a of the solutions of another problem. total computable function f such that x ∈ A iff f (x) ∈ B for all x.
There are several inequivalent ways of reducing a problem to another We shall write A ≤m B or more explicitly f : A ≤m B. problem. If f is injective, then we are talking about one-one reducibility . The differences between different reductions consists in the manner and extent to which information about B is allowed to settle questions about A.
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 4/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 5/64 Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set Examples Elementary Properties
1. K is m-reducible to {x | φx = 0}, {x | c ∈ Wx} and {x | φx is total }. Let A, B, C be sets.
0 if x ∈ Wx y if x ∈ Wx 1. ≤m is reflexive: A ≤m A. f (x, y) = fN(x, y) = 0 ↑ if x ∈ W ↑ if x ∈ W x x f : A ≤m A is the identity function. B 2. Rice Theorem is proved by showing that K ≤m {x | φx ∈ }. 2. ≤m transitive: A ≤m B, B ≤m C ⇒ A ≤m C. B g(y) if x ∈ Wx x ∈ Wx ⇒ φk(x) = g ∈ Let f : A ≤m B, g : B ≤m C, then g ◦ f : A ≤m C. fg(x, y) = ↑ if x ∈ Wx x ∈ Wx ⇒ φk(x) = f∅ ∈ B 3. A ≤m B iff A ≤m B. 3. {x | φx is total } ≤ m {x | φx = 0}. If g : A ≤m B, then x ∈ A ⇔ f (x) ∈ B; so x ∈ A ⇔ g(x) ∈ B. φk(x) = 0 ◦ φx Hence g : A ≤m B.
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 6/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 7/64
Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set Elementary Properties (2) Elementary Properties (3)
4. If A is recursive and B ≤m A, then B is recursive. 7. (i). A ≤m N iff A = N; (ii). A ≤m ∅ iff A = ∅. g : B ≤m A; then cB(x) = cA(g(x)) . So cB is computable. (i).“ ⇐": By reflexivity, N ≤m N. (i).“ ⇒": Let f : A ≤m N, then x ∈ A ⇔ f (x) ∈ N. Thus A = N. 5. If A is recursive and B = ∅, N, then A ≤m B. (ii). A ≤m ∅ ⇔ A ≤m N ⇔ A = N ⇔ A = ∅. b if x ∈ A; Let b ∈ B, c ∈ B, f (x) = ; then f is c if x ∈ A. 8. (i). N ≤m A iff A = ∅; (ii). ∅ ≤m A iff A = N. computable. (i). “ ⇒": Let f : N ≤ A, then A = Ran (f ), so A = ∅ (f is x ∈ A ⇔ f (x) ∈ B. m total). (i). “ ⇐": If A = ∅, choose c ∈ A. If g(x) = c, we have 6. If A is r.e. and B ≤m A, then B is r.e. g : N ≤m A. C Let g : B ≤m A, A = Dom (h), ( h ∈ 1); then B = Dom (h ◦ g) (ii). ∅ ≤m A ⇔ N ≤m A ⇔ A = ∅ ⇔ A = N. (B is r.e.)
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 8/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 9/64 Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set Corollary Corollary (2)
Fact . If A is r.e. and is not recursive, then A ≤ m A and A ≤ m A. Corollary . Neither {x | φx is total } nor {x | φx is not total } is m-reducible to K. Proof . “ A ≤ m A": By contradiction, if A ≤m A, then A is r.e., then A is recursive! Proof . By contradiction, if {x | φx is total } ≤ m K, and K is r.e., then {x | φx is total } is r.e. (same as {x | φx is not total }). “A ≤ m A": By contradiction, if A ≤m A, then A ≤m A, then A is recursive! However, by Rice-Shapiro Theorem, Neither {x | φx is total } nor {x | φ is not total } is r.e. x Notation: It contradicts to our intuition that A and A are equally difficult.
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 10/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 11/64
Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set Theorem Many-One Equivalence
Theorem . A is r.e. iff A ≤m K. Definition . Two sets A, B are many-one equivalent , notation A ≡m B (abbreviated m-equivalent), if A ≤m B and B ≤m A. Proof . “ ⇐". Since A ≤m K, and K is r.e., then A is r.e. 1, if x ∈ A, Suppose A is r.e. Let f (x, y) be f (x, y) = ↑, if x ∈/ A. Theorem . ≡m is an equivalence relation.
By s-m-n Theorem ∃s(x) : N → N such that f (x, y) = φs(x)(y). Proof . (1). Reflexivity: A ≤m A ⇒ A ≡m A. It is clear that x ∈ A iff φs(x)(s(x)) is defined iff s(x) ∈ K. Hence A ≤m K. (2). Symmetry: A ≡m B ⇒ B ≤m A, A ≤m B ⇒ B ≡m A. (3). Transitivity: A ≡m B, B ≡m C ⇒ A ≤m C, C ≤m A ⇒ A ≡m C. Notation . K is the most difficult partially decidable problem.
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 12/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 13/64 Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set Examples Example (2)
1. {x | c ∈ Wx} ≡ m K. 3. If A is r.e. but not recursive, then A ≡ m A. y if x ∈ Wx “⇐": fN(x, y) = ⇒ K ≤m {x | c ∈ Wx} A is r.e. but not recursive ⇒ A ≤ m A, A ≤ m A. ↑ if x ∈ Wx “⇒": {x | c ∈ Wx} is r.e., so {x | c ∈ Wx} ≤ m K. 4. {x | φx = 0} ≡ m {x | φx is total }. Thus {x | c ∈ Wx} ≡ m K. “⇐": φk(x) = 0 ◦ φx ⇒ { x | φx is total } ≤ m {x | φx = 0}. 0 if φ (y) = 0; 2. If A is recursive, A = ∅, N, then A ≡m A. x “⇒": Let φk(x)(y) = . ↑ if φx(y) = 0. A = ∅, N ⇒ A = ∅, N. Then φx = 0 ⇔ φk(x) is total ⇒ { x | φx = 0} ≤ m {x | φx is total }. A is recursive, by previous theorem A ≤m A. Similarly, A ≤m A.
CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 14/64 CSC363-ComputabilityTheory@SJTU XiaofengGao ReducibilityandDegree 15/64
Reduction and Degree Many-One Reduction Reduction and Degree Many-One Reduction Relative Computability Degrees Relative Computability Degrees Turing Reducibility m-Complete r.e. Set Turing Reducibility m-Complete r.e. Set m-Degree Expression
Definition . The set of m-degrees is ranged over by a, b, c,... .
Definition . Let dm(A) be {B | A ≡m B}. Definition (Partial Order on m-Degree) . Let a, b be m-degrees.
Definition . An m-degree is an equivalence class of sets under the (1). a ≤m b iff A ≤m B for some A ∈ a and B ∈ b. relation ≡m. It is any class of sets of the form dm(A) for some set A. (2). a