Shaded similarity matrix can be naturally used in nearest
neighbor visualization since they both use the concept of
similarity or distance. When using shaded similarity matrix
in nearest neighbor, we do not need to consider the seriation.
The objects within the same classes are put together. We do
not care about the order of objects in the same class. That
is, they are ordered randomly within the same class.
We use a distinct color to represent a different class. The
shade of a color indicates the distance or similarity. The
larger the distance, the lighter the shade. Since each cell
represents the distance between two objects, what if the two
objects belongs to different classes? We just set the color
to one class. This is not a problem because the similarity
matrix is usually symmetrical.