By component plane representation we can visualize the relative component distributions of the input data. Component plane representation can be thought as a sliced version of the Self-Organizing Map. Each component plane has the relative distribution of one data vector component. In this representation, dark values represent relatively small values while white values represent relatively large values. By comparing component planes we can see if two components correlate. If the outlook is similar, the components strongly correlate.
Figure 2.10: The component plane representation of the SOM
This is a clear visualization of correlation between the vector components. For example, there is correlation between the components (j), (k) and (l), for example. By picking a same neuron in each plane (in the same location), we could assemble the relative values of a codebook vector of the network.