This rule is necessary in order to do 'one-step-recall' simulations. For more informations see [Ama89]. For the calculation the bias following facts are assumed:

- The implemented net is an autoassociative net

The neurons of an autoassociative net have to be input and output neurons at the same time. A Hopfield network would be an example for such a net. - Fixed number of 1s

The patterns which are to be saved have a fixed number of 1s.

The parameter h1 and h2 are required. Where h1 is the number of ones per
pattern and h2 is the probable degree of distortion in percent. The parameters
have to be inserted in field1 and field2.
This initialization function should be used only in connection with the
` Hopfield_Fixed_Act` update function. As mentioned in section
the ` Hebb_FixAct` algorithm is a learning algorithm.

Niels.Mache@informatik.uni-stuttgart.de

Tue Nov 28 10:30:44 MET 1995