Empirical scaling of the length of the longest increasing subsequences of random walks

Abstract: We provide Monte Carlo estimates of the scaling of the length $L_{n}$ of the longest increasing subsequences of $n$-steps random walks for several different distributions of step lengths, short and heavy-tailed. Our simulations indicate that, barring possible logarithmic corrections, $L_{n} \sim n^{\theta}$ with the leading scaling exponent $0.60 \lesssim \theta \lesssim 0.69$ for the heavy-tailed distributions of step lengths examined, with values increasing as the distribution becomes more heavy-tailed, and $\theta \simeq 0.57$ for distributions of finite variance, irrespective of the particular distribution. The results are consistent with existing rigorous bounds for $\theta$, although in a somewhat surprising manner. For random walks with step lengths of finite variance, we conjecture that the correct asymptotic behavior of $L_{n}$ is given by $\sqrt{n}\ln n$, and also propose the form of the subleading asymptotics. The distribution of $L_{n}$ was found to follow a simple scaling form with scaling functions that vary with $\theta$. Accordingly, when the step lengths are of finite variance they seem to be universal. The nature of this scaling remains unclear, since we lack a working model, microscopic or hydrodynamic, for the behavior of the length of the longest increasing subsequences of random walks.