Title :
Fast self-organizing of n-dimensional topology maps
Author :
Haese, Karin ; vom Stein, Heinz-Dieter
Author_Institution :
University of the Federal Armed Forces, Hamburg, Signal Processing Holstenhofweg 85, D - 22043 Hamburg
Abstract :
The self-organizing algorithm, proposed in this contribution, is more efficient than the original one, because it starts at a coarse lattice and refines the lattice of the map using spline interpolation at well determined learning steps until a quantization criteria is reached. Therefore the feature map becomes self-growing. The proposed algorithm also includes a hierarchical search for the nearest neighbour. All these enhancements lead to a time complexity of order O(logN) of the self-organizing algorithm for a nA-dimensional map with N neurons in each dimension.
Keywords :
Interpolation; Lattices; Neurons; Quantization (signal); Splines (mathematics); Time complexity;
Conference_Titel :
European Signal Processing Conference, 1996. EUSIPCO 1996. 8th
Conference_Location :
Trieste, Italy
Print_ISBN :
978-888-6179-83-6