Consider, for example, the definition for $\Sigma_2^p$ complexity class.

$$ x \in L \Leftrightarrow \exists u_1 \forall u_2 \;M(x, u_1, u_2) = 1, $$

where $u_1, u_2 \in \{0,1\}^{p(|x|)}$, for some polynomial $p$. Here, $M$ must be polynomial time. But polynomial in the size of what exactly? For example, if we choose (guess) some $u_1$, do I consider it to be fixed size when talking about time complexity of $M$? More precisely, should $M$ be polynomial only in the size of $x$?

An example. Consider the problem whether, given a graph $A$, there exists a graph $B$ such that $B$ is subgraph isomorphic to $A$.

$$A \in L \Leftrightarrow \exists B \; \text{SubGraphIsomorphic}(A, B) = 1 $$

Now, subgraph isomorphism is NP-complete. If $B$ is fixed, then there is a TM that implements $\text{SubGraphIsomorphic}$ in deterministic polynomial time. If $B$ is not fixed, then I cannot claim such a thing unless I know $\sf P=NP$. Is this problem in $\Sigma_{1}^{p}$, i.e. $\sf NP$? (Ok, this problem has trivial solutions, but I hope it helps to pinpoint my confusion.)

My confusion generalizes for all $\Sigma_{i}^p$.

没有正确的解决方案

许可以下: CC-BY-SA归因
不隶属于 cs.stackexchange
scroll top