Another, quite interesting option is to use another SOM to cluster the map units. This kind of structure is often referred to as a hierarchical SOM. Usully a ``hierarchical SOM'' refers to a tree of maps the lower of which act as a preprocessing stage to higher ones. As the hierarchy is traversed upwards, the information becomes more and more abstract. Hierarchical self-organizing networks were first proposed by Luttrell [25]. He pointed out that although adding extra layers to a vector quantizer yields a higher distortion in reconstruction, it also effectively reduces the complexity of the task. Another advantage is that different kinds of representations are available from different levels of the hierarchy.
A multilayer hierarchical SOM (HSOM) for clustering was introduced by
Lampinen and Oja [24]. In the HSOM the BMU of an input
vector 
is sought from the first-layer map and its index is
given as input to the second-layer map. If more than one data vector
is fed to the first layer map the whole data histogram can be given to
the second layer instead of a single index. This kind of approach
have been used e.g. in arranging document databases [23].