...mining

Not all authors use the same terminology; data mining can also be used to refer to whole process, even as a synonym of KDD.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...below

Ripley (1996) divides statistical data-analysis methods into clustering methods, projection methods, and multidimensional scaling (MDS) methods. Differentiation between the latter two is not useful here, since for patterns represented as vectors in a Euclidean space the MDS methods essentially form nonlinear projections.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...essentially

The algorithm Hastie and Stuetzle (1989) proposed for finding discretized principal curves resembles the batch version [Kohonen, 1995c] of the SOM algorithm, although the details are different.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...allow

Increase in the map size brings more resolution into the mapping. The ``stiffness'' of the map, or smoothness of the mapping, can be controlled independently of the map size by changing the final width of the learning neighborhood.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.