A Modified Visualization-Model of High-dimensional Data
一种改进的高维数据可视化模型

The Visualization-Induced Self-Organizing Maps (ViSOM), as one of the artificial neural networks models, has been successfully applied in the analysis of visualization of high-dimensional data. However, it has two weaknesses. Firstly, it does not consider the correlation of data. Secondly, much memory will be used up if the output nodes are too large, and contrarily, the visibility results of data will be difficult to be analyzed if the output nodes are too small. In order to overcome the above two weaknesses of ViSOM, a modified algorithm named MViSOM, based on ViSOM, as well as a visualization-model of high-dimensional data, based on ICA (Independent Component Analysis)and MViSOM, are proposed in this paper. Finally, the experiments also show that IMViSOM method has advantages over ViSOM because of its excellent classified effect of swarm data and high calculating speed, confirming the correctness and reasonableness for the proposed model in this paper.