Hierarchical growing cell structures

Abstract

We propose a hierarchical self-organizing neural network ("HiGS") with adaptivearchitecture and simple topological organization. This network combines features ofFritzke's Growing Cell Structures and traditional hierarchical clustering algorithms.The height and width of the tree structure depend on the user-specified level of errordesired, and the weights in upper layers of the network do not change in later phases ofthe learning algorithm. Parameters such as node deletion rate are...