A novel associative memory system based on sharp-angled hat spline functions

The learning problem of associative memory system (AMS) is investigated from the viewpoint of the approximation of input-to-output mapping function. In order to improve the generalization properties in learning multidimensional mappings of polynomial functions with order no more than 2, this paper proposes a novel associative memory system (AMS) based on a set of sharp-angled hat spline functions (SHSFs). The approximation capabilities of conventional AMS is evaluated at first, then a novel SHSF and the general structure of SHSF-AMS with corresponding interpolation and training algorithms are presented. Pertinent numerical simulations have shown the effectiveness of the proposed method.