A Distributed Online Sequential Learning Algorithm Based on Zero Gradient Sum Strategy

In this paper, a distributed online sequential zero-gradient-sum algorithm is proposed, which is based on the discretetime zero-gradient-sum algorithm for directed networks. It also integrates concepts from the online sequential learning algorithm, allowing each node to perform online learning while maintaining the necessary initialization conditions for iterative updates. By introducing auxiliary variables, each node performs online learning while ensuring that the learned node parameters satisfy the initialization conditions of the ZGS iterations. Consequently, the parameters of all nodes converge to a global optimal solution through distributed iterations. This paper provides a theoretical proof of the algorithm and validates its performance through simulation experiments.