arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.01534cs.ITmath.ITmath.STstat.TH
收藏

条件正态性与有限状态维度再探

Conditional normality and finite-state dimensions revisited

LIRMM, Univ Montpellier, CNRS(法国国家科学研究中心蒙彼利埃大学拉格朗日数学、信息与技术实验室)

Alexander Shen

AI总结:

本文重新审视相对有限状态维度与条件正态性,通过推广无条件正态性的复杂度和赌博刻画,简洁证明分组频率与赌博方法的等价性并解决遗留问题。

摘要:

正态比特序列的概念由 Borel 于 1909 年提出;它是个体随机对象的第一个定义。正态性是一种较弱的随机性概念,只要求任意长度 n 的全部 $2^n$ 个因子(子串)都以相同的极限频率 $2^{-n}$ 出现。后来人们提出了许多更强的随机性定义,在这一背景下,正态性被定位为“对抗有限记忆对手的随机性”。有限状态可压缩性的一种定量度量(即有限状态维度)也被引入,而正态性意味着有限状态维度达到最大值(等于 1)。最近,Nandakumar、Pulari 和 S(2023)针对一个二元序列相对于另一个二元序列(后者被视为 oracle)的情形,提出了相对有限状态维度的概念,以及相应的条件(相对)正态性概念。(此前已有不同的条件随机性概念得到研究,但并非针对有限记忆情形。)他们建立了条件正态性与有限状态维度在分组频率方法和赌博方法之间的等价性。在本短文中,我们重新审视他们的定义,并说明如何通过推广关于(无条件)正态性和维度的已知刻画——即可压缩性(有限状态复杂度)、超加性复杂度度量以及赌博(有限状态 gale)——而轻松获得这种等价性,从而也回答了上述论文中遗留的一些问题。

英文摘要:

The notion of a normal bit sequence was introduced by Borel in 1909; it was the first definition of an individual random object. Normality is a weak notion of randomness requiring only that all $2^n$ factors (substrings) of arbitrary length~$n$ appear with the same limit frequency $2^{-n}$. Later many stronger definitions of randomness were introduced, and in this context normality found its place as ``randomness against a finite-memory adversary''. A quantitative measure of finite-state compressibility was also introduced (the finite-state dimension) and normality means that the finite state dimension is maximal (equals~$1$). Recently Nandakumar, Pulari and S (2023) introduced the notion of relative finite-state dimension for a binary sequence with respect to some other binary sequence (treated as an oracle), and the corresponding notion of conditional (relative) normality. (Different notions of conditional randomness were considered before, but not for the finite memory case.) They establish equivalence between the block frequency and the gambling approaches to conditional normality and finite-state dimensions. In this note we revisit their definitions and explain how this equivalence can be obtained easily by generalizing known characterizations of (unconditional) normality and dimension in terms of compressibility (finite-state complexity), superadditive complexity measures and gambling (finite-state gales), thus also answering some questions left open in the above-mentioned paper.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑