1

I think I can't understand the meaning of language being decidable. The next case makes no sense to me:

Considering I have language L1 which is regular, and language L2 which is in RE\R (in particular, it cannot be decided).

The right-division of any regular language with some language is regular, and in particular in R.

so:

L1/L2={x ∈ Σ* | ∃ y ∈ L2 : xy ∈ L1}

is decidable.

But I can't see why. I can't describe an algorithm that decides L1\L2, so how come this language is decidable? I know how to define the transition function, but does it means that the language is decidable? After all, I can also define a transition function to the accepting problem, it still does not make it decidable.

After all, to check if x belongs to L1 \ L2 I need to go over words and check if there is a word in L2 that completes x to a word in L1, but if there is none, I will run forever.

I think I'm missing something, so I'm very confused. Would appreciate help!

Ella
  • 109
  • 6

1 Answers1

3

In fact, if $L_1$ is regular then $L_1/L_2$ is regular for any $L_2$.

Indeed, consider a DFA for $L_1$ with transition function $\delta$ and accepting states $F$. We modify the set of accepting states to $$ F' = \{ q : \delta(q,w) \in F \text{ for some } w \in L_2 \}. $$

You might object that "$F'$ cannot be computed". So what? We never said that you can compute a DFA for $L_1/L_2$ given one for $L_1$ and some reasonable representation of $L_2$. All we claim is that $L_1/L_2$ is regular.

Check out this question for a similar but simpler situation.

Yuval Filmus
  • 276,994
  • 27
  • 311
  • 503
  • But if it can be computed, I can also define a function that decides problems like the halting problem and claim that the halting problem is decidable. (I have read the question you added but I still can’t understand) – Ella Jun 14 '20 at 04:42
  • I don’t see how you can solve the halting problem this way. The issue is that there are infinitely many possible Turing machines. – Yuval Filmus Jun 14 '20 at 06:01