Q-learning based algorithm for small-world network generation

Abstract

Small-world networks are an important type of complex network structures with small average shortest path lengths and large average clustering coefficients. A variety of generation algorithms for small-world networks have been given in existing studies, but less attention has been paid to how to optimize the small-world property of generative networks. In this paper, considering multiple objective optimization problems, small-worldness is defined to evaluate the small-world properties of generative networks, and Q-learning-based algorithms for small-world network generation are proposed and compared with WS small-world networks. Compared with the WS small-world network, the network generated by the Q-learning algorithm performs better in each metric.