Scribd
Upload
54 views21 pages

Reductions and Rice's Theorems: Deepak D'Souza

This document discusses reductions and Rice's theorems regarding the properties of recursively enumerable (r.e.) languages. It defines reductions between languages and shows that if one language reduces to another, then their recursive and r.e. properties are preserved. Rice's theorems state that any non-trivial property of r.e. languages is undecidable, and any non-monotone property is not recursively enumerable. Proofs are provided using reductions from the halting problem and its complement. Applications include showing that properties like "accepts the empty language" are undecidable.

Uploaded by

deepakk
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
54 views21 pages

Reductions and Rice's Theorems: Deepak D'Souza

This document discusses reductions and Rice's theorems regarding the properties of recursively enumerable (r.e.) languages. It defines reductions between languages and shows that if one language reduces to another, then their recursive and r.e. properties are preserved. Rice's theorems state that any non-trivial property of r.e. languages is undecidable, and any non-monotone property is not recursively enumerable. Proofs are provided using reductions from the halting problem and its complement. Applications include showing that properties like "accepts the empty language" are undecidable.

Uploaded by

deepakk
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

Reductions Rice’s theorems

Reductions and Rice’s theorems

Deepak D’Souza

Department of Computer Science and Automation


Indian Institute of Science, Bangalore.

20 November 2019
Reductions Rice’s theorems

Outline

1 Reductions

2 Rice’s theorems
Reductions Rice’s theorems

Reductions

Let L ⊆ A∗ and M ⊆ B ∗ be two languages. We say L reduces to


M and write L ≤ M iff there exists a computable map
σ : A∗ → B ∗ such that

w ∈ L iff σ(w ) ∈ M.
A∗ B∗

σ
L
M
Reductions Rice’s theorems

Examples of reductions

Let L be the language {n | n is even } (with say n encoded in


binary). Let L0 be the language {l#m#r | l mod m = r }.
Then L ≤ L0 via the computable map n 7→ n#2#0.
Does L0 reduce to L?
Let L be the language {M | M accepts }. Then

HP ≤ L.

Describe a computable map σ which witnesses the reduction.


Reductions Rice’s theorems

Reductions and recursive/re-ness

Theorem
If L ≤ M then:
1 If M is r.e. then so is L.
2 If M is recursive then so is L.
Or to put it differently:
Theorem
If L ≤ M then:
1 If L is not r.e. then neither is M.
2 If L is not recursive then neither is M.
Reductions Rice’s theorems

Examples of reductions

Let L be the language {M | M accepts }. Then

HP ≤ L.

Describe a computable map σ which witnesses the reduction.


Hence, since HP is undecidable (i.e. not recursive) so is L.
Reductions Rice’s theorems

Examples of reductions

Let L be the language {M | M accepts a regular language}. Then

¬HP ≤ L.

Describe a computable map σ which witnesses the reduction.


Hence, since ¬HP is undecidable (i.e. not recursive) so is L.
In fact, since ¬HP is not r.e., we can say that L is not r.e..
Reductions Rice’s theorems

Rice’s theorem

Theorem (Rice)
Any non-trivial property of r.e. languages is undecidable.
Reductions Rice’s theorems

Rice’s theorem

Theorem (Rice)
Any non-trivial property of r.e. languages is undecidable.

Theorem (Rice)
Any non-monotone property of r.e. languages is not even
recursively enumerable.
Reductions Rice’s theorems

Properties of languages

A property P of languages over an alphabet A is a subset of


languages over A.
Languages over A

RE languages
P Property P
Reductions Rice’s theorems

Non-trivial and montone properties

A property P is a non-trivial property of r.e. languages, if


there is at least one r.e. language L satisfying P, and another
L0 not satisfying P.
Reductions Rice’s theorems

Non-trivial and montone properties

A property P is a non-trivial property of r.e. languages, if


there is at least one r.e. language L satisfying P, and another
L0 not satisfying P.
E.g. “is empty” is non-trivial
Reductions Rice’s theorems

Non-trivial and montone properties

A property P is a non-trivial property of r.e. languages, if


there is at least one r.e. language L satisfying P, and another
L0 not satisfying P.
E.g. “is empty” is non-trivial
“is not accepted by a TM” is trivial.
A property P of languages is monotone (w.r.t r.e. languages)
if for all r.e. sets A and B, whenever A ⊆ B and P(A), we
have P(B).
In other words, P is monotone if whenever a set has the
property P, all its supersets have it as well.
Reductions Rice’s theorems

Non-trivial and montone properties

A property P is a non-trivial property of r.e. languages, if


there is at least one r.e. language L satisfying P, and another
L0 not satisfying P.
E.g. “is empty” is non-trivial
“is not accepted by a TM” is trivial.
A property P of languages is monotone (w.r.t r.e. languages)
if for all r.e. sets A and B, whenever A ⊆ B and P(A), we
have P(B).
In other words, P is monotone if whenever a set has the
property P, all its supersets have it as well.
“is infinite” is monotone,
Reductions Rice’s theorems

Non-trivial and montone properties

A property P is a non-trivial property of r.e. languages, if


there is at least one r.e. language L satisfying P, and another
L0 not satisfying P.
E.g. “is empty” is non-trivial
“is not accepted by a TM” is trivial.
A property P of languages is monotone (w.r.t r.e. languages)
if for all r.e. sets A and B, whenever A ⊆ B and P(A), we
have P(B).
In other words, P is monotone if whenever a set has the
property P, all its supersets have it as well.
“is infinite” is monotone,
“L(M) is finite” is not monotone.
Reductions Rice’s theorems

Rice’s theorems
For a property P, we define

LP = {M | L(M) satisfies P}.

Theorem (Rice 1953)


Any non-trivial property of r.e. languages is undecidable. That is,
if P is a non-trivial property of r.e. languages, then the language
LP is not recursive.

Theorem (Rice 1956)


Any non-monotone property of r.e. languages is not even
recursively enumerable. That is, if P is a non-monotone property
of r.e. languages, then the language LP is not even recursively
enumerable.
Reductions Rice’s theorems

Proof of Rice’s Theorem 1

Let P be a non-trivial property of r.e. languages. Then there


are TM’s K and T such that L(K ) satisfies P and L(T ) does
not satisfy P.
We show that LP = {M | L(M) satisfies P} is not recursive.
Case 1: If ∅ does not satisfy P. We reduce HP to LP .
Given M#x, construct a machine M 0 = σ(M#x) that on
input y
saves y on a separate track
writes x on its tape
runs as M on input x
if M halts on x, M 0 runs as K on y and accepts iff K accepts.


0 L(K ) if M halts on x
L(M ) =
∅ if M does not halt on x.
Reductions Rice’s theorems

Proof of Rice’s Theorem 1

Case 2: If ∅ satisfies P. We reduce ¬HP to LP .


Given M#x, construct a machine M 0 = σ(M#x) that on
input y
saves y on a separate track
writes x on its tape
runs as M on input x
if M halts on x, M 0 runs as T on y and accepts iff T accepts.


0 ∅ if M does not halt on x
L(M ) =
L(T ) if M halts on x.
Reductions Rice’s theorems

Proof of Rice’s Theorem 2

Let P be a non-monotone property of r.e. sets.


Then there are TM’s K and T such that L(K ) ⊆ L(T ) and
L(K ) satisifes P but L(T ) does not.
We show ¬HP ≤ LP .
Given M#x output the description of M 0 that
Given input y on Tape 1.
Copies y on Tape 2, writes x on Tape 3
Run (in an interleaved fashion) as M on x, K on y , and T on
y.
accept iff either
K accepts y , or,
M halts on x and T accepts y .
Reductions Rice’s theorems

Proof of Rice’s Theorem 2

Notice that:

0 L(K ) if M does not halt on x
L(M ) =
L(T ) if M halts on x.
Reductions Rice’s theorems

Some applications

From Rice’s Theorem 1:


“Accepts ” is undecidable.
“Accepts an infinite language” is undecidable.

{M | M accepts an infinite language}.

From Rice’s Theorem 2:


“Accepts the empty language” is “highly” undecidable
(non-r.e.).
“Accepts a finite language” is highly undecidable (non-r.e.).

{M | M accepts a finite language}.

“Accepts a regular language” is highly undecidable (non-r.e.).

You might also like