The first step of this function is to determine the interval from which the weights for the initialization tasks are selected. Depending upon the parameters and , which have to be provided in field1 and field2, the interval could be:
The chosen weight vector for every neuron j in the Kohonen layer consists of the weights , which are uniformly values from the above described intervals. Where are the weights of the links from neuron i to neuron j. The length of is < 1. The last step is to normalize the length of to 1. All weights of the neurons of the Grossberg layer are random values from the intervals mentioned above. See also chapter for some additional informations.