Scaling exponents of step-reinforced random walks (Q2660391)

Let \((\varepsilon_n)_{n\geq 1}\) be a \(0-1\)-valued deterministic sequence with \(\varepsilon_1=1\); \(X_1\), \(X_2,\ldots\) independent identically distributed real-valued random variables; \(U(2)\), \(U(3),\ldots\) independent random variables such that \(U(n)\) has a discrete uniform distribution on \(\{1,\ldots, n-1\}\), the two sequences being independent. Let \((S_n)_{n\geq 1}\) be the standard random walk with jumps \(X_k\) and assume that the distribution of \(X_1\) belongs to the normal domain of attraction of an \(\alpha\)-stable distribution, \(\nu_\alpha\) say, for \(\alpha\in (0,2]\). The latter means that \(\lim_{n\to\infty}n^{-1/\alpha}S_n=Y_\alpha\) in distribution, where \(Y_\alpha\) is a random variable with distribution \(\nu_\alpha\). The parameter \(\alpha\) is known as scaling exponent. Further, consider a perturbed random walk \((\hat S_n)_{n\geq 1}\) that the author calls a step-reinforced random walk defined by \(\hat S_n:=\hat X_1+\ldots+\hat X_n\) for \(n\in\mathbb{N}\), where \(\hat X_n:=X_{\varepsilon_1+\ldots+\varepsilon_n}\) if \(\varepsilon_n=1\) and \(\hat X_n:=\hat X_{U(n)}\) if \(\varepsilon_n=0\). The purpose of the paper is to find out how the reinforcement affects the distributional growth of partial sums. To exemplify the author's findings, assume that \(\sum_{n\geq 1}n^{-2}|\varepsilon_1+\ldots+\varepsilon_n-qn|<\infty\) for some \(q\in (0,1)\) and that \(q>1-1/\alpha\) when \(\alpha>1\). Then the scaling exponents of \(S\) and \(\hat S\) coincide and are equal to \(\alpha\). If \(q<1-1/\alpha\), then the scaling exponent of \(\hat S\) is \((1-q)^{-1}<\alpha\). The case in which \(\varepsilon_1+\cdots+\varepsilon_n\) exhibits a sublinear growth is also analyzed. The main results of the paper are accompanied with appealing intuitive explanations. The paper finishes with a number of interesting remarks which particularly suggest several directions for future work.