A New Training Method for Large Self Organizing Maps
Neural Processing Letters
| Hi-index | 0.00 |
SOM neural network is one of the most commonly used Clustering algorithom in the text clustering. The initial connection weights of SOM neural network will affect the degree of convergence. If the Initial connection weights are not set appropriate, that will cause in a long wandering around the local minimum, accordingly lower the speed of convergence, or even cause local convergence or not convergence. Initializing the connection weights closer to the center of each category can highten the speed of convergence.Because text-data-intensive area may contain category center or close to category center , this paper presents a hierarchical clustering method to detect text-data-intensive areas and use the center of the K detected text-data-intensive areas to initialize the connection weight of SOM neural network, in order to improve the speed of SOM neural network convergence. The experimental results showed that: ensuring the effectiveness of text clustering, the text clustering speed is greatly improved.