Abstract

We present a new self-organizing neural network model having two variants. The first variant performs unsupervised learning and can be used for data visualization, clustering, and vector quantization. The main advantage over existing approaches, e.g., the Kohonen feature map, is the ability of the model to automatically find a suitable network structure and size. This is achieved through a controlled growth process which also includes occasional removal of units. The second variant of the model is a supervised learning method which results from the combination of the abovementioned self-organizing network with the radial basis function (RBF) approach. In this model it is possible - in contrast to earlier approaches - to perform the positioning of the RBF units and the supervised training of the weights in parallel. Therefore, the current classification error can be used to determine where to insert new RBF units. This leads to small networks which generalize very well. Results on the t...

Citations

2726	The self-organizing map - Kohonen - 1990
936	Self-Organized Formation of Topologically Correct Feature Maps - Kohonen - 1982
599	The perceptron: A probabilistic model for information storage and organization - Rosenblatt - 1958
590	The cascade-correlation learning architecture - Fahlman, Lebiere - 1990
118	Learning to tell two spirals apart - Lang, Witbrock - 1988
115	Learning with localized receptive fields - Moody, Darken - 1988
108	A neural-gas network learns topologies - Martinetz, Schulten
88	The “neural” phonetic typewriter - Kohonen - 1988
63	Dynamic error propagating networks - Robinson - 1994
45	R.: Incremental grid growing: Encoding high-dimensional structure into a two-dimensional feature map - Blackmore, Mükkulainen - 1993
37	LEDA: A library of efficient data types and algorithms - Mehlhorn, Naher - 1989
35	Learning with the Self-Organizing Map - Ritter
31	Constructing Hidden Units Using Examples and Queries - Baum, Lang - 1990
25	Speaker Normalisation for Automatic Speech Recognition - Deterding - 1989
24	Kohonen feature maps and Growing Cell Structures – A - Fritzke - 1992
21	Phonotopic Maps: Insightful Representation of Phonologic Features for Speech Recognition - Kohonen, Makisara, et al. - 1984
19	A study of the application of Kohonen-type neural networks to the traveling salesman problem - Favata, Walker - 1991
12	Improving the learning speed in topological maps of pattern - Rodrigues, Almeida - 1990
9	Vector quantization with a growing and splitting elastic net - Fritzke - 1993
5	CMU benchmark collection for neural net learning algorithms - Fahlman - 1993
4	A fully neural approach to image compression - Schweizer, Parladori, et al. - 1991
3	von der Malsburg (1976), How Patterned Neural Connections Can Be Set Up by Self-Organization - Willshaw
1	A neural network which adapts its structure to a given set of patterns - Jokusch - 1990