Approximations for nonlinear functions
Irwin W. Sandberg · IEEE Transactions on Circuits and Systems I Fundamental Theory and Applications · 1992
Results concerning the uniform approximation of nonlinear functionals on compact sets are discussed. The results are of interest in connection with neural network-like classifiers for continuous-time signals and approximations for the input-output maps of dynamic systems. It is shown that function-space feedforward neural networks with one input layer of bounded linear functionals and one hidden nonlinear layer each are universal approximators of real continuous functionals on compact subsets of a normed linear space.>