将日志空间与多项式时间分开


24

显然,在确定性对数空间()中可确定的任何问题最多都在多项式时间()中运行。和之间有很多复杂性类。示例包括,,,,,。人们普遍认为,。LPLPNLLogCFLNCiSACiACiSCiLP

在我的一篇博客文章中,我提到了两种证明方法(以及相应的猜想)。这两种方法都基于分支程序,并且相隔20年!是否有其他方法和/或猜想来分离LPLP分开(或)将和之间的任何中间类别分开。LP


觉得这个问题一个TM运行顺序压缩是相关
VZN

Answers:


21

电路深度下限(等效地,式尺寸下限)可能是最自然的方法:的super log2(n)深度下界中的问题P会分离PL和Karchmer-Wigderson通信复杂度技术可以为此自然而然。


3
自然证明障碍在这里不会成为问题吗?我很好奇为什么会这样。
Suresh Venkat 2010年

6
是的,这样的证据肯定看起来必须是“非自然的”,但据我所知,它必须是博客文章中提到的其他方法。
诺姆

8

[1]证明了mincost-flow实例的下限,其实例的位大小与图的大小相比足够大(但仍是线性的),并且进一步证明了如果对于足够小的输入,可以显示相同的下限比特大小意味着(因此PL)。从总体上讲,这与Noam的回答相同,因为它是关于证明电路深度的下界(=公式大小的下界),但似乎是与Karchmer-Wigderson游戏截然不同的方向。PNCPL

更详细地,[1]显示以下内容。使用与本文相同的表示法,让表示最小成本流语言。我们可以将n顶点图上的mincost-flow语言表示为L n ,看作是对于k n = Θ n 2Z k n 的子集,其中整数由位串编码。令B a n 表示Z k n 中所有向量的集合LnL(n)Zk(n)k(n)=Θ(n2)B(a,n)Zk(n)其中每个整数坐标的位大小最多为。给定一个函数f x 1x k(我们将在后面指定哪种函数),如果L n ∩中的点points ,则说fB a n 内将L n 分隔开ñ 正是那些Xanf(x1,,xk)fL(n)B(a,n)L(n)B(a,n)使得 f x= 1xB(a,n)f(x)=1

命题[1,命题7.3]如果B a n )中det M x分隔,其中是大小的矩阵,其项是(线性组合,使得,然后。L(n)B(a,n)det(M(x))M2n/dx1,,xka<1/(2d)PNC

这里,位绑定的和大小绑定的关系至关重要。在同一篇论文中,他展示了:an2n/d

定理[1,定理7.4]对于所有足够大的位界,前述命题的假设成立。a

上述定理的证明使用了一些沉重的铁锤作为黑匣子,但除此之外还是基本的(请注意:“基本” “ 简单 ”)。也就是说,它使用实半代数变体的连接分量数上的Milnor-Thom边界(Ben-Or使用相同的边界来证明真实计算树模型中元素的唯一性/排序的下界),柯林斯分解(用来证明对),一般位置参数和其他一些想法的有效量词消除。但是,所有这些技术仅取决于所涉及的多项式的阶数,因此不能像上述命题一样用于证明(实际上是[1,Prop。7.5]构造一个多项式RPNCg of the same degree as det such that the above proposition fails with g in place of det). Analyzing this situation and looking for properties that went beyond degree was one of the inspirations for GCT.

[1] K. Mulmuley. Lower Bounds in a Parallel Model without Bit Operations. SIAM J. Comput., 28(4), 1460–1509, 1999


8

It made my day when my friend James told me that this thread from long ago was rekindled. Thank you for that.

Also, I had an urge to share some interesting references that are relevant to L vs Log(DCFL) vs Log(CFL). Have a great day!

http://link.springer.com/chapter/10.1007%2F978-3-642-14031-0_35#page-1

http://link.springer.com/chapter/10.1007/3-540-10003-2_89?no-access=true

http://link.springer.com/chapter/10.1007%2F978-3-642-00982-2_42#page-1

http://www.researchgate.net/publication/220115950_A_Hardest_Language_Recognized_by_Two-Way_Nondeterministic_Pushdown_Automata


7

this new paper was just highlighted by Luca Aceto in his blog as an EATCS best student paper at ICALP 2014 & has a novel way of separating NL/P:

  • Hardness Results for Intersection Non-Emptiness Wehar

    We carefully reexamine a construction of Karakostas, Lipton, and Viglas (2003) to show that the intersection non-emptiness problem for DFA's (deterministic finite automata) characterizes the complexity class NL. In particular, if restricted to a binary work tape alphabet, then there exist constants c1 and c2 such that for every k intersection non-emptiness for k DFA's is solvable in c1klog(n) space, but is not solvable in c2klog(n) space. We optimize the construction to show for an arbitrary number of DFA's intersection non-emptiness is not solvable in o(nlog(n)log(log(n))) space. Furthermore, if there exists a function f(k)=o(k) such that for every k intersection non-emptiness for k DFA's is solvable in nf(k) time, then P≠NL. If there does not exist a constant c such that for every k intersection non-emptiness for k DFA's is solvable in nc time, then P does not contain any space complexity class larger than NL.

By using our site, you acknowledge that you have read and understand our Cookie Policy and Privacy Policy.
Licensed under cc by-sa 3.0 with attribution required.