Winner-Relaxing Self-Organizing Maps

TL;DR

The paper introduces the Winner-Relaxing Kohonen Algorithm, a generalized self-organizing map that allows precise control over magnification, achieving optimal information-theoretic mapping with minimal computational overhead.

Contribution

It presents a new family of self-organizing maps with adjustable magnification, extending previous variants and enabling optimal mappings in one dimension.

Findings

Magnification exponent of 4/7 for the original variant

Generalized version allows exponent tuning from 1/2 to 1

Algorithm requires minimal extra computations

Abstract

A new family of self-organizing maps, the Winner-Relaxing Kohonen Algorithm, is introduced as a generalization of a variant given by Kohonen in 1991. The magnification behaviour is calculated analytically. For the original variant a magnification exponent of 4/7 is derived; the generalized version allows to steer the magnification in the wide range from exponent 1/2 to 1 in the one-dimensional case, thus provides optimal mapping in the sense of information theory. The Winner Relaxing Algorithm requires minimal extra computations per learning step and is conveniently easy to implement.