A memetic-based algorithm for solving stable extension in abstract argumentation frameworks

In abstract argumentation frameworks, stable semantics is one of the important criteria for evaluating the acceptability of arguments, and the computation of stable extensions is a core problem in argumentation models. Existing methods for computing stable extensions can generally be divided into reduction-based methods and direct solving methods. These two types of methods often face challenges such as high computational complexity and expanding search spaces when dealing with large-scale random instances. This paper proposes a stable extension solving method based on a hybrid evolutionary algorithm (Memetic Algorithm For Stable Extension, abbreviated as Mase). This method combines heuristic methods, tabu search algorithms, and genetic algorithms, introducing tabu lists and an adaptive parent crossover strategy to achieve iterative optimization of stable extensions. In the experiment, more than 80 benchmark datasets are randomly generated using the AFGen generator and compared with algorithms such as Haywood and u-toksia. The results show that Mase consumes significantly less time in solving stable extensions for large-scale random instances compared to other methods. In addition, ablation experiments are designed by removing the tabu strategy and the adaptive parent crossover strategy, creating two ablation versions. The results show that Mase outperforms both ablation versions in all instances, verifying the collaborative improvement of the tabu strategy and the adaptive parent crossover strategy on the overall performance of the algorithm.