Dynamic two-dimensional feature map with modified ant clustering for data visualization

摘要

The self-organizing networks are useful tools suitable for data analysis in which networks learn the topology of the complicated high-dimensional data structure and can project it on the two-dimensional plane. However, it has been noticed that those methods does not always visualize the high-dimensional clusters adequately on the lower-dimensional plane. In this paper, we proposed the visualization method by the ant clustering algorithm in order to construct the structures on the two-dimensional grid spare as feature map.