论文信息 - Evolutionary self-organizing map

Evolutionary self-organizing map

Extending Kohonen's SOM the paper proposes one kind of dynamically growing neural network, called evolutionary SOM (ESOM). Firstly, the output layer of the SOM is represented by a so-called neighborhood graph, where nodes are neurons' weights and edges are the neighborhood relationships of SOM. Then two basic differentiation operations, node differentiation and edge differentiation, are proposed for network differentiation. As in nature's evolution, each generation of ESOM includes several species of neural nets and the survivors of competition will differentiate to the next generation. This kind of evolution is implemented as two new modules of ESOM, in addition to Kohonen's SOM toolbox in Matlab. A cross pattern with 1000 data points is taken as example. The results show that there are a large quantity of unnecessary neurons in Kohonen's SOMs; whereas the resultant ESOM has much smaller size and better fitness to training input.

Maiga Chang | Jia-Sheng Heh | Horng-Jyh Yu

[1] R.M. Gray,et al. A greedy tree growing algorithm for the design of variable rate vector quantizers [image compression] , 1991, IEEE Trans. Signal Process..

[2] Teuvo Kohonen,et al. Self-Organization and Associative Memory , 1988 .

[3] Helge Ritter,et al. Extending Kohonens Self-Organizing Mapping Algorithm to Learn Ballistic Movements , 1988 .

[4] Klaus Schulten,et al. Topology-conserving maps for learning visuo-motor-coordination , 1989, Neural Networks.

[5] Liang-Gee Chen,et al. Vector quantization using tree-structured self-organizing feature maps , 1994, IEEE J. Sel. Areas Commun..

[6] Richard Lippmann,et al. Neural Net and Traditional Classifiers , 1987, NIPS.

[7] Jorma Laaksonen,et al. Variants of self-organizing maps , 1990, International 1989 Joint Conference on Neural Networks.

[8] Helge J. Ritter,et al. Three-dimensional neural net for learning visuomotor coordination of a robot arm , 1990, IEEE Trans. Neural Networks.

[9] Thomas Villmann,et al. Growing a hypercubical output space in a self-organizing feature map , 1997, IEEE Trans. Neural Networks.

[10] Dieter A. Mlynski,et al. Mapping and hierarchical self-organizing neural networks for VLSI placement , 1997, IEEE Trans. Neural Networks.