C02: Zeroing neural network (ZNN)

Paper Submission for C02

SMC2018:C02 submission site (external site)

Abstracts

Recently, a special type of recurrent neural networks called zeroing neural network (ZNN) has received great achievements in time-varying problems solving. It differs considerably from conventional gradient-based neural networks in terms of the problem to be solved, error function, design formula, dynamic equation, and the utilization of time-derivative information. ZNN is viewed as a systematic solution to time-varying problems, and represents a summit in the research on artificial neural networks. From ZNN design point of view, the residual errors computed by ZNN models can globally and exponentially converge to zero. Also, any monotonically increasing odd activation function can be constructed and exploited in ZNN design, thereby establishing ZNN models with more excellent properties, such as finite-time convergence and noise-tolerant capability. In consideration of the significant advantages of ZNN, this special session seeks to promote new research investigations in ZNN and related areas (in particular the research on ZNN with disturbance). Following such a session, we expect to shift the ZNN research from study in ideal situation to that with theoretical consideration of non-ideal working environments.

Session Chairs