Effects of Backtracking on PageRank

TL;DR

This paper explores three variations of PageRank that modify backtracking probabilities, analyzing their equivalence on certain graph types and comparing their effectiveness in clustering tasks.

Contribution

It introduces and compares non-backtracking, μ-, and ∞-PageRank variants, revealing their equivalence on specific graph classes and assessing their clustering performance.

Findings

Standard PageRank and variants are equivalent on regular and bipartite biregular graphs.

Different PageRank variants exhibit distinct clustering capabilities.

The study provides insights into how backtracking adjustments affect centrality measures.

Abstract

In this paper, we consider three variations on standard PageRank: Non-backtracking PageRank, $\mu$-PageRank, and $\infty$-PageRank, all of which alter the standard formula by adjusting the likelihood of backtracking in the algorithm's random walk. We show that in the case of regular and bipartite biregular graphs, standard PageRank and its variants are equivalent. We also compare each centrality measure and investigate their clustering capabilities.