1
$\begingroup$

Given a function $f : \Sigma^* \to \mathbb{N}$, define function $f_{-1}$ as: $f_{-1}(x) = f(x) - 1$ if $f(x) > 0$, and $f_{-1}(x) = 0$ otherwise. Moreover, say that a class ${\cal C}$ of functions is closed under decrement if for every $f \in {\cal C}$, it holds that $f_{-1} \in {\cal C}$.

Define $\#\text{PE}$ [2,3] as the class of functions in $\#\text{P}$ with easy decision versions, that is, the class of functions $f \in \#\text{P}$ for which there exists a polynomial-time deterministic TM $M$ such that $L(M) = \{ x \in \Sigma^* \mid f(x) > 0 \}$. It was shown in [1] that if $\#\text{P}$ is closed under decrement, then $\text{NP} \subseteq \text{SPP}$. But the ideas in the proof of this result cannot be used to provide evidence that $\#\text{PE}$ is not closed under decrement. So I was wondering what is known about this problem. Is $\#\text{PE}$ closed under decrement? Can at least one prove that if $f \in \#\text{PE}$, then $f_{-1} \in \#\text{P}$? Any help with this would be greatly appreciate it.

[1] Mitsunori Ogiwara, Lane A. Hemachandra: A Complexity Theory for Feasible Closure Properties. J. Comput. Syst. Sci. 46(3): 295-325 (1993)

[2] Aris Pagourtzis. On the complexity of Hard Counting Problems with Easy Decision Version. In Proceedings of the 3rd Panhellenic Logic Symposium, 2001.

[3] Aris Pagourtzis, Stathis Zachos: The Complexity of Counting Functions with Easy Decision Version. MFCS 2006: 741-752

$\endgroup$
6
$\begingroup$

If $\text{#PE}$ is closed under decrement then $\text P=\text{NP}$: consider $\text{#SAT}+1$. Conversely if $\text P=\text{NP}$ then $\text{#PE}=\text{#P}$ is closed under decrement because we can always find an accepting path if one exists then ignore it.

I don't know whether $f\in\text{#PE}$ always implies $f_{-1}\in\text{#P}$. Your $M$ might not provide an accepting path for us to ignore.

$\endgroup$
  • $\begingroup$ I do not understand your proof that #PE is closed under decrement implies P=NP. What do you mean by consider #SAT + 1? $\endgroup$ – T.... Jun 13 '17 at 14:51
  • $\begingroup$ @Turbo: I mean a language like SAT but with one extra accepting path. For example the witnesses could be $(t,y_1,...,y_n)$ where either: ($t=0=y_1=\dots=y_n=0$) or ($t=1$ and $(y_1,...y_n)$ is a witness for the input SAT formula). $\endgroup$ – Colin McQuillan Jun 23 '17 at 13:01
  • $\begingroup$ @ColnMcQuillan Ok I see it. Since $\#SAT_{new}=\#SAT_{given} + 1$ is in $\#PE$ ($SAT_{new}$ is new $SAT$ instance by adding an accepting path to $SAT_{given}$ and this always puts $SAT_{new}$ in $YES$ instance and so decision is in $P$) then if $\#PE$ is closed we have $\#SAT_{given}=\#SAT_{new}-1$ in $\#PE$ which implies $SAT_{given}$ is in $P$ and since $SAT_{given}$ is arbitrary so $P=NP$. is this what you intend? That is a neat non-constructive existence proof. $\endgroup$ – T.... Jun 24 '17 at 4:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.