The Kohonen learning algorithm does not use the activation functions of the units. Therefore it is basically unimportant which activation function is used in the SOM. For display and evaluation reasons, however, the two activation functions Act_Component and Act_Euclid have been implemented in SNNS. Act_Euclid copies the Euclidian distance of the unit from the training pattern to the unit activation, while Act_Component writes the weight to one specific component unit into the activation of the unit.