In this paper, a novel shape-tunable feedforward neural network is pro
posed. Based on the steepest descent method, an autotuning algorithm t
hat enables the proposed neural network to possess the ability of auto
matic shape-tuning is derived Due to the ability of auto-shaping, the
flexibility and nonlinearity capacity of the neural network is increas
ed significantly. Furthermore. the novel feature of automatic shaping
prevents the nonlinear neurons from saturation, and therefore the scal
ing procedure. which is usually unavoidable for the traditional fixed-
shape neural networks, becomes unnecessary. Simulation results indicat
e that the proposed shape-tunable neural network gives better agreemen
t than the traditional fixed-shape one does, even though fewer nodes a
re used. Moreover, the convergence properties are more superior. To de
monstrate the capability of the proposed shape-autotuning neural netwo
rks to a great extent. we adopted it as a learning-type direct control
ler. Some related problems were studied. Copyright (C) 1996 Elsevier S
cience Ltd