Fast and asymptotic computation of the fixation probability for Moran processes on graphs

TL;DR

This paper introduces a fast, asymptotic method for computing fixation probabilities in Moran processes on graphs, especially bipartite graphs, by using embedded Markov chains to improve computational efficiency over traditional Monte Carlo simulations.

Contribution

It develops an asymptotic approach using embedded Markov chains to efficiently compute fixation probabilities on complex graphs, extending previous theoretical results.

Findings

The method provides faster computation of fixation probabilities.

Comparison shows the approach outperforms Monte Carlo in certain network types.

The paper revisits key theorems on star graphs and clarifies recent results.

Abstract

Evolutionary dynamics has been classically studied for homogeneous populations, but now there is a growing interest in the non-homogenous case. One of the most important models has been proposed by Lieberman, Hauert and Nowak, adapting to a weighted directed graph the classical process described by Moran. The Markov chain associated with the graph can be modified by erasing all non-trivial loops in its state space, obtaining the so-called Embedded Markov chain (EMC). The fixation probability remains unchanged, but the expected time to absorption (fixation or extinction) is reduced. In this paper, we shall use this idea to compute asymptotically the average fixation probability for complete bipartite graphs. To this end, we firstly review some recent results on evolutionary dynamics on graphs trying to clarify some points. We also revisit the 'Star Theorem' proved by Lieberman, Hauert and Nowak for the star graphs. Theoretically, EMC techniques allow fast computation of the fixation probability, but in practice this is not always true. Thus, in the last part of the paper, we compare this algorithm with the standard Monte Carlo method for some kind of complex networks.