arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2007.00381cond-mat.stat-mechmath-phmath.MP
收藏

更新次数分布的极限形式

Limit Forms of the Distribution of the Number of Renewals

Stanislav Burov

摘要:

在这项工作中的渐近性质的$Q_t(N)$,更新($N$),发生在时间$t$的概率。虽然分布在很长时间的形式,即$t\to\infty$,是非常有名的,并与高斯中心极限定理或列维稳定的法律,替代限制大量的更新,即$N\to\infty$,是少得多的注意。我们解决这个限制的大$N$,并发现它达到了一个普遍的形式,完全依赖于分析性质的更新时间的分布。给出了$Q_t(N)$的显式公式,并给出了有限$N$的修正和收敛于普遍渐近极限的必要条件.我们的结果表明,$N/t$的大偏差率函数存在,并达到一个普遍的线性增长(对数校正)的$N/t\to\infty$限制。无论平均更新时间的存在或幂律统计的存在,这一结果成立。

英文摘要:

In this work the asymptotic properties of $Q_t(N)$ ,the probability of the number of renewals ($N$), that occur during time $t$ are explored. While the forms of the distribution at very long times, i.e. $t\to\infty$, are very well known and are related to the Gaussian Central Limit Theorem or the Lévy stable laws, the alternative limit of large number of renewals, i.e. $N\to\infty$, is much less noted. We address this limit of large $N$ and find that it attains a universal form that solely depends on the analytic properties of the distribution of renewal times. Explicit formulas for $Q_t(N)$ are provided, together with corrections for finite $N$ and the necessary conditions for convergence to the universal asymptotic limit. Our results show that the Large Deviations rate function for $N/t$ exists and attains an universal linear growth (up to logarithmic corrections) in the $N/t\to\infty$ limit. This result holds irrespective of the existence of mean renewal time or presence of power-law statistics.

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。

↑