Investigation of alternative strategies and quality measures for controlling the growth process of the growing hierarchical self-organizing map

The self-organizing map (SOM) is a very popular neural network model for data analysis and visualization of high-dimensional input data. The growing hierarchical self-organizing map (GHSOM) - being one of the many architectures based on the SOM - has the property of dynamically adapting its architecture during training by map growth as well as creating a hierarchical structure of maps, thus reflecting hierarchical relations in the data. This allows for viewing portions of the data at different levels of granularity. We review different SOM quality measures and also investigate alternative strategies as candidates for guiding the growth process of the GHSOM in order to improve the hierarchical representation of the data.