A Variable Parameter Zeroing Neural Network for Resolving Time-Variant Quadratic Minimization with Preferable Performance

The variable parameter recurrent neural network has powerful abilities to resolve the various time-variant problems. In this work, a new-type neural network model to handle the time-variant quadratic minimization issue is proposed and researched. Distinct from the generalized original zeroing neural network (OZNN) model, the transfer function part of the proposed variable parameter zeroing neural network (VPZNN) model use the sign-bi-power function, and the parameter part use the specially constructed time-variant parameter. Finally, numerical simulation is carried out with and without noise. The simulation results exhibit that the proposed VPZNN model has preferable convergence capability and robustness than the OZNN model and the finite-time Zhang neural network (FTZNN) model.