Let $L \subseteq \Sigma^{\star}$ and let $M^{\Sigma^{\star} \times \Sigma^{\star}}(\{0,1\})$ an infinite matrix such that for each $x,y\in \Sigma^\star$: $$ m_{x,y}=\begin{cases} 1 & x y\in L\\ 0 & \text{otherwise} \end{cases} $$
I need to show that if $L\in REG$ then there is a finite number of distinct rows in $M$.
$M$ rows resemble me of binary representations of real numbers in $[0,1]$ so perhaps I need to find a function $f:M_i \to [0,1] $ and conclude that since a regular language has a cardinality of $\aleph_0$ there can be only a finite number of distinct rows, but:
- I don't know if that is really the direction to prove this claim.
- I can't find such an $f$. Furthermore, what if $M$'s rows are the unit vectors? There are infinite many distinct rows but the similarity to real numbers binary representation is no longer intuitive to me.
Thanks for your tips!
This is part of the Myhill-Nerode theorem: the rows for $x$ and $y$ are identical if and only if for all $z\in\Sigma^*$, $xz\in L$ iff $yz\in L$ iff there is no distinguishing extension for $x$ and $y$. There is a proof of the theorem at the link, but you might want to try to find one yourself from this HINT: the number of distinct rows is equal to the number of states in the smallest DFA recognizing $L$. (In other words, think in terms of the operation of a DFA recognizing $L$.)