Intelligent training of a multi-resolution associative memory network

Abstract

New challenges in US Navy tactical vehicle applications require flexible adaptive structures for real-time intelligent system identification and control tasks. Due to its speed of computation and linear learning behavior, the architecture of lattice-based associative memory networks is particularly attractive. The main problem of these networks is the curse of dimensionality, which means that the memory requirement is exponentially dependent on the input dimension. In this paper, a novel structure called multi-resolution associative memory (MAM) network is proposed to overcome the curse of dimensionality. Preliminary simulations and real-time experiments are performed using this structure, and the results are discussed. A fuzzy rule-bare system is also proposed as a method to implement the heuristic that is at the heart of the MAM network's training algorithm.