Adaptive spline neural network for signal processing applications

In this paper, we study the properties of a new kind of real and complex domain artificial neural networks called adaptive spline neural networks (ASNN), which are able to adapt their activation functions by varying the control points of a Catmull-Rom cubic spline. Most of all, we are interested in generalization capability and we can show that this architecture can be seen as a sub-optimal realization of the additive spline based model obtained by the reguralization theory. This new kind of neural network can be implemented as a very simple structure being able to improve the generalization capabilities using few training epochs. Due to its low architectural complexity this network can be used to cope with several nonlinear DSP problem at high throughput rate.