One-Sided Approximation and Interpolation Operators Generating Hyperbolic Sigma-Pi Neural Networks
Burkhard Lenze · Birkhäuser Basel eBooks · 1997
In this paper, we show how to design three-layer feedforward neural networks with hyperbolic sigma-pi units in the hidden layer in order to act as one-sided approximation and interpolation devices for regular gridded data. We obtain the concrete networks in real-time using a one-shot learning scheme based on special approximation operators which are generated by sampling the given discrete information on a regular grid. In this context, it is essential that we do not require any smoothness conditions regarding the underlying data function f . At the end of the paper we briefly discuss an application of our strategy. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.