On decision-boundary-based approaches for structure selection of multilayered neural nets
D. Chen, Phillip Burrell · 2002
The optimal structure design for the multilayer feedforward neural network (MFNN) based classifier is of a fundamental importance in the neural network research area. In this paper several typical decision-boundary-based methods for the optimal structure design of the MFNN-based classifier are discussed. The main characteristics of each approach are analyzed and compared. Finally, future research directions are outlined.