Gaussian Neighborhood Function for the Self-Organizing Map

Given the following variables:

Let d be the distance between the best output node and the output node whose weights are being modified.
Let c be the current radius (neighborhood size)

Then, for each output node, the scaling factor is:
e^{-d² / (2c²)}
Multiply each weight update by this scaling factor. There is no longer any need to check neighborhood membership; the Gaussian reduces the impact exponentially as the radius shrinks.