Constructed by a set of distributively associated subnets

a multi-valued learning network （KLN） is proposed in this paper for approximating nonlinear function. Based on the proposed netword

the multivalued logic control is further studied. Network operations are mainly integer addition and logic judgment. Simulation results show that the KLN-based multi-valued control system is strong in self-learning and simple in computation

and has good learning control performance for complex nonlinear systems.