Distributed Event-Triggered Learning for Stochastic Configuration Networks
Xiaowen Ge, Wu Ai, Yujun Zhou · 2020
The purpose of this paper is to improve a distributed learning algorithm for stochastic configuration networks (SCNs) with event-triggered communication mechanism. During this process, the centralized SCNs problem will be converted to a distributed problem. The solution of this problem is the discrete-time zero-gradient-sum (ZGS) strategy. Then lead into the event-triggered communication mechanism to reduce the communication complexity of this distributed problem. With this mechanism, a trigger condition that whether the error exceeds the threshold or not determines whether or not each agent communicates. Information are transmitted between the agents only when they are in extraordinary need and the information is exchanged asynchronously. And based on ZGS strategy, we proposes a distributed learning algorithm. This algorithm can be used to train SCNs with event-triggered communication over networks. Finally, the output results of one simulation example are given to prove the validity of the proposed algorithm.