Social behavior study under pervasive social networking based on decentralized deep reinforcement learning

Pervasive social networking (PSN) provides instant social activities such as communications and gaming, which attracts a growing attention. Under this circumstance, the study and analysis of users' behaviors has a profound meaning, which may be extended to other relevant fields. In this article, we define and quantify users' patterns to study social behaviors, after discussing and reconsidering social characteristics in PSN. Meanwhile, we treat PSN as a market, based on the standpoint that data can be priced and tradable. After analyzing its market structure, we describe PSN as a monopolistically competitive market, which contains multiple providers selling specialized goods. Afterwards, we study the social behaviors in this market from an economic perspective, namely applying market models. Finally, decentralized deep reinforcement learning is proposed to estimate users' patterns and to solve market models, the prisoner's dilemma and the Cournot model to be specific. Simulation results demonstrate the flexibility and efficiency of our algorithms.