On Explicit Constructions of Extremely Depth Robust Graphs
Authors:
Jeremiah Blocki,
Mike Cinkoske,
Seunghoon Lee,
Jin Young Son
Abstract:
A directed acyclic graph $G=(V,E)$ is said to be $(e,d)$-depth robust if for every subset $S \subseteq V$ of $|S| \leq e$ nodes the graph $G-S$ still contains a directed path of length $d$. If the graph is $(e,d)$-depth-robust for any $e,d$ such that $e+d \leq (1-ε)|V|$ then the graph is said to be $ε$-extreme depth-robust. In the field of cryptography, (extremely) depth-robust graphs with low ind…
▽ More
A directed acyclic graph $G=(V,E)$ is said to be $(e,d)$-depth robust if for every subset $S \subseteq V$ of $|S| \leq e$ nodes the graph $G-S$ still contains a directed path of length $d$. If the graph is $(e,d)$-depth-robust for any $e,d$ such that $e+d \leq (1-ε)|V|$ then the graph is said to be $ε$-extreme depth-robust. In the field of cryptography, (extremely) depth-robust graphs with low indegree have found numerous applications including the design of side-channel resistant Memory-Hard Functions, Proofs of Space and Replication, and in the design of Computationally Relaxed Locally Correctable Codes. In these applications, it is desirable to ensure the graphs are locally navigable, i.e., there is an efficient algorithm $\mathsf{GetParents}$ running in time $\mathrm{polylog} |V|$ which takes as input a node $v \in V$ and returns the set of $v$'s parents. We give the first explicit construction of locally navigable $ε$-extreme depth-robust graphs with indegree $O(\log |V|)$. Previous constructions of $ε$-extreme depth-robust graphs either had indegree $\tildeω(\log^2 |V|)$ or were not explicit.
△ Less
Submitted 22 March, 2022; v1 submitted 8 October, 2021;
originally announced October 2021.
On the Speed of an Excited Asymmetric Random Walk
Authors:
Mike Cinkoske,
Joe Jackson,
Claire Plunkett
Abstract:
An excited random walk is a non-Markovian extension of the simple random walk, in which the walk's behavior at time $n$ is impacted by the path it has taken up to time $n$. The properties of an excited random walk are more difficult to investigate than those of a simple random walk. For example, the limiting speed of an excited random walk is either zero or unknown depending on its initial conditi…
▽ More
An excited random walk is a non-Markovian extension of the simple random walk, in which the walk's behavior at time $n$ is impacted by the path it has taken up to time $n$. The properties of an excited random walk are more difficult to investigate than those of a simple random walk. For example, the limiting speed of an excited random walk is either zero or unknown depending on its initial conditions. While its limiting speed is unknown in most cases, the qualitative behavior of an excited random walk is largely determined by a parameter $δ$ which can be computed explicitly. Despite this, it is known that the limiting speed cannot be written as a function of $δ$. We offer a new proof of this fact, and use techniques from this proof to further investigate the relationship between $δ$ and speed. We also generalize the standard excited random walk by introducing a "bias" to the right, and call this generalization an excited asymmetric random walk. Under certain initial conditions we are able to compute an explicit formula for the limiting speed of an excited asymmetric random walk.
△ Less
Submitted 4 September, 2017; v1 submitted 28 August, 2017;
originally announced August 2017.
