Fernández Domingos Elias, Santos Francisco C, Lenaerts Tom
Machine Learning Group, ULB, Campus la Plaine, Brussels 1050, Belgium.
AI Lab, VUB, Pleinlaan 9, Brussels 1050, Belgium.
iScience. 2023 Mar 17;26(4):106419. doi: 10.1016/j.isci.2023.106419. eCollection 2023 Apr 21.
Evolutionary Game Theory (EGT) provides an important framework to study collective behavior. It combines ideas from evolutionary biology and population dynamics with the game theoretical modeling of strategic interactions. Its importance is highlighted by the numerous high level publications that have enriched different fields, ranging from biology to social sciences, in many decades. Nevertheless, there has been no open source library that provided easy, and efficient, access to these methods and models. Here, we introduce EGTtools, an efficient hybrid C++/Python library which provides fast implementations of both analytical and numerical EGT methods. EGTtools is able to analytically evaluate a system based on the replicator dynamics. It is also able to evaluate any EGT problem resorting to finite populations and large-scale Markov processes. Finally, it resorts to C++ and MonteCarlo simulations to estimate many important indicators, such as stationary or strategy distributions. We illustrate all these methodologies with concrete examples and analysis.
进化博弈论(EGT)为研究集体行为提供了一个重要框架。它将进化生物学和种群动态学的思想与战略互动的博弈论建模相结合。数十年来,众多高水平的出版物丰富了从生物学到社会科学等不同领域,凸显了其重要性。然而,一直没有一个开源库能方便、高效地使用这些方法和模型。在此,我们介绍EGTtools,一个高效的C++/Python混合库,它提供了分析和数值EGT方法的快速实现。EGTtools能够基于复制者动态对系统进行分析评估。它还能够借助有限种群和大规模马尔可夫过程评估任何EGT问题。最后,它采用C++和蒙特卡罗模拟来估计许多重要指标,如稳态或策略分布。我们用具体的例子和分析来说明所有这些方法。