New Heuristic of Self Organizing Map Using Updating Distribution

Self organizing map (SOM) is a special type of artificial neural networks for unsupervised learning. SOM has been used as a good tool of clustering. Generally the weights of SOM are updated by a learning process which is depended on not distributions but values. After complete updating process, the final weights are determined by fixed values. So, the clustering result from a complete updating is only outcome. But, for example, the cognitive behaviors of human being are shown different results from the given experience. In this paper, we propose a new heuristic of SOM (NHSOM) using updating distributions. NHSOM is able to provide diverse results from the weight distributions. In our experimental results, we verify efficient and improved performances of NHSOM to compare other competitive algorithms using the data sets from UCI machine learning repository and synthesis.