Abstract | ||
---|---|---|
Genetic algorithm is introduced to network optimization to overcome the limitation of conventional SOM network. Based on this idea, a new model of structural adapting self-organizing neural network is proposed. In this model, each neuron is regarded as individual of evolutionary population and three operators are constructed as follows:growing operator, pruning operator and stochastic creating operator. In the algorithm, the accumulative error of neuron is selected as fitness function each iteration, and the neurons on compete layer are generated or deleted adaptively according to the values of fitness function until there is not any change of neuron on compete layer. Simulation experiments indicate that this structural adaptive network has better performance than conventional SOM network. |
Year | DOI | Venue |
---|---|---|
2007 | 10.1007/978-3-540-72393-6_109 | ISNN (2) |
Keywords | Field | DocType |
structural adaptive network,structural adapting self-organizing maps,neural network,genetic algorithm,fitness function,new model,network optimization,structural adapting,accumulative error,pruning operator,conventional som network,simulation experiment,self organization | Population,Computer science,Stochastic neural network,Fitness function,Probabilistic neural network,Self-organizing map,Time delay neural network,Artificial intelligence,Artificial neural network,Genetic algorithm,Machine learning | Conference |
Volume | ISSN | Citations |
4492 | 0302-9743 | 1 |
PageRank | References | Authors |
0.36 | 7 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xin-zheng Xu | 1 | 219 | 14.45 |
Wenhua Zeng | 2 | 136 | 14.83 |
Zuopeng Zhao | 3 | 14 | 1.39 |