Scribd Logo
Upload

Welcome to Scribd!

  • 214 views

    Proof of Myhill Nerode Theorem of Toc

    Uploaded by

    Abhishek Raj
    The document summarizes the Myhill-Nerode theorem, which provides a necessary and sufficient condition for a language to be regular. The theorem states that a language is regular if and only if it has a finite index. The index of a language is the size of the largest set of strings that are pairwise distinguishable by the language. The proof of the theorem relies on two lemmas: 1) if a language is recognized by a DFA with n states, its index is ≤ n, and 2) if a language has a finite index n, there exists a DFA with n states that recognizes the language.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd

    Proof of Myhill Nerode Theorem of Toc

    Uploaded by

    Abhishek Raj
    214 views3 pages
    The document summarizes the Myhill-Nerode theorem, which provides a necessary and sufficient condition for a language to be regular. The theorem states that a language is regular if and only if it has a finite index. The index of a language is the size of the largest set of strings that are pairwise distinguishable by the language. The proof of the theorem relies on two lemmas: 1) if a language is recognized by a DFA with n states, its index is ≤ n, and 2) if a language has a finite index n, there exists a DFA with n states that recognizes the language.

    Original Description:

    proof with examples

    Original Title

    proof of myhill nerode theorem of toc

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document summarizes the Myhill-Nerode theorem, which provides a necessary and sufficient condition for a language to be regular. The theorem states that a language is regular if and only if it has a finite index. The index of a language is the size of the largest set of strings that are pairwise distinguishable by the language. The proof of the theorem relies on two lemmas: 1) if a language is recognized by a DFA with n states, its index is ≤ n, and 2) if a language has a finite index n, there exists a DFA with n states that recognizes the language.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    214 views3 pages

    Proof of Myhill Nerode Theorem of Toc

    Uploaded by

    Abhishek Raj
    The document summarizes the Myhill-Nerode theorem, which provides a necessary and sufficient condition for a language to be regular. The theorem states that a language is regular if and only if it has a finite index. The index of a language is the size of the largest set of strings that are pairwise distinguishable by the language. The proof of the theorem relies on two lemmas: 1) if a language is recognized by a DFA with n states, its index is ≤ n, and 2) if a language has a finite index n, there exists a DFA with n states that recognizes the language.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    CS 4510 : Automata and Complexity : Spring 2010

    Myhill-Nerode Theorem
    Subrahmanyam Kalyanasundaram
    January 26, 2010
    Myhill-Nerode theorem gives a necessary and sucient condition for a language to be regular. The
    pumping lemma gives only a necessary condition for a language to be regular, and Myhill-Nerode
    theorem does one better in that respect. The theorem was proved by John Myhill and Anil Nerode
    in 1958. We state the necessary denitions and the theorem below, and explain the proof in a
    top-down manner.
    Denition 1 (Distinguishable by a language). Let r, j be strings and 1 be a language over the
    same alphabet . r, j are said to be distinguishable by 1 if .

    such that r. 1 and j. , 1


    or vice versa.
    Also, if r and j are not distinguishable by 1, they are said to be indistinguishable by 1. This is
    denoted by r

    j.
    Exercise 1
    Show that

    is an equivalence relation.
    Denition 2 (Pairwise Distinguishable). Let 1 be a language and A be a set of strings. A is
    pairwise distinguishable by 1 if every two distinct strings r, j A are distinguishable by 1.
    Denition 3 (Index). The index of a language 1 is the size of the largest set A of strings which
    is pairwise distinguishable by 1.
    These are the denitions we need. The theorem is stated below.
    Theorem 1 (Myhill-Nerode Theorem). Language 1 is regular if and only if it has a nite index.
    Moreover, the index of 1 is equal to the size of the smallest DFA which recognizes 1.
    The proof of the theorem follows by the following two lemmas.
    Lemma 1. If 1 is recognized by a DFA with / states, then index(1) /.
    Lemma 2. If index(1) = / < , then there exists a DFA with / states that recognizes 1.
    The proof of the theorem follows very easily once we assume the above two lemmas.
    1
    Proof of Myhill-Nerode Theorem. We have by denition, 1 is regular there exists a DFA which
    recognizes 1.
    Suppose 1 is regular. Choose the smallest (least number of states) DFA that recognizes 1. Let /
    be the number of states of this DFA. By lemma 1, index(1) /. In particular
    index(1) Size of the smallest DFA that recognizes 1
    This also implies that 1 has nite index as well. This shows one direction of the theorem.
    For the other direction, let index(1) = / < . Then by lemma 2 there exists a DFA with / states
    that recognizes 1. This shows the other direction of the theorem, that 1 is regular. This also shows
    that
    Size of the smallest DFA that recognizes 1 index(1)
    By the above two inequalities, we can conclude that index(1) = the size of the smallest DFA that
    recognizes 1.
    Now let us see the proofs of the lemmas. Note that we use the notation c(, r) for strings r

    .
    c(, r) denotes the state that the DFA reaches starting from after reading the string r.
    Proof of Lemma 1. We prove this by contradiction. We have a language 1 recognized by a DFA
    with / states. We assume, for the sake of contradiction, that index(1) /. This means that there
    is a set A, A /, such that A is pairwise distinguishable by 1.
    Let ` be the DFA with / states that recognizes 1. Let
    0
    be the start state of `. We have A /.
    Then by pigeonhole principle, there exist two distinct r, j A such that c(
    0
    , r) = c(
    0
    , j).
    Now we shall assert that r, j are pairwise indistinguishable, thus contradicting our assumption that
    index(1) /. Consider any .

    .
    c(
    0
    , r.) = c(c(
    0
    , r), .) = c(c(
    0
    , j), .) = c(
    0
    , j.)
    So r. 1 j. 1. r, j are pairwise indistinguishable, and hence by contradiction, index(1)
    /.
    Proof of Lemma 2. Suppose index(1) = /. We have to show the existence of a DFA with / states
    that recognizes 1. We shall in fact, construct such a DFA. Let A = {r
    1
    , r
    2
    , . . . , r

    } be a largest
    set which is pairwise distinguishable by 1 (the existence of A follows by the denition of index).
    We build DFA ` = (Q, , c,
    0
    , 1). is chosen to be the same as the alphabet of 1. We choose
    Q = {
    1
    ,
    2
    , . . . ,

    }. It would be useful to think of each

    as corresponding to the string r

    A.
    For each o , we dene c(

    , o) as follows. Consider the string r

    o. We have r

    for some
    r

    A. This follows from the fact that A is a largest pairwise distinguishable set by 1. If r

    for any r

    , this would mean that A {r

    o} would be a bigger pairwise distinguishable set by 1.


    Now dene c(

    , o) =

    .
    Similarly, we have the empty string -

    for some

    . Set

    =
    0
    as the starting state.
    We claim (and shall prove soon) that c(

    , n) =

    .
    Dene 1 = {

    1}. We have dened all the components of the DFA ` now. All that remains
    is to show that this DFA recognizes 1. Suppose r 1. Then r

    for some r

    , such that r

    1
    2
    (Why?). By the above claim,
    r = -r

    = c(
    0
    , r) =

    But

    1 since r

    1. So r is recognized by `.
    If r , 1, then r

    for some r

    , 1. Like above, we get that c(


    0
    , r) =

    . Again

    , 1 since
    r

    , 1. So in this case ` does not recognize r. So ` recognizes exactly the language 1.


    The only part that we are yet to prove is the claim above.
    Proof of Claim. The claim is that c(

    , n) =

    .
    We prove this by induction on n, the length of n.
    When n = 0, n = -. In this case c(

    , n) = c(

    , -) =

    . The claim is true since r

    - = r

    .
    When n = 1, the claim is true by the denition of the transition function c.
    When n 1, we use induction. Suppose the claim holds for all n such that n |. Let n
    be such that n = | + 1. Let n = o where = | and o = 1, o . We have
    c(

    , n) = c(

    , o) = c(c(

    , ), o) = c(

    1
    , o) =

    2
    where

    1
    = c(

    , ) and

    2
    = c(

    1
    , o). This means that r

    and r

    2
    o (this
    follows by induction hypothesis). The claim is true since
    r

    n = r

    1
    o

    2
    3

    You might also like