In order to work with various neural network models and learning algorithms, different initialization functions that initialize the components of a net are required. Backpropagation, for example, will not work properly if all weights are initialized to the same value.

To select an initialization function, one must click in the ` INIT` line of the control panel.

The following initialization functions are available:

All these functions receive their input from the five init parameter fields in the control panel. See figure

Here is a short description of the different initialization functions:

- ART1_Weights
- ART2_Weights
- ARTMAP_Weights
- CC_Weights
- ClippHebb
- CPN_Rand_Pat
- CPN_Weights_v3.2
- CPN_Weights_v3.3
- DLVQ_Weights
- Hebb
- Hebb_FixAct
- JE_Weights
- Kohonen_Weights_v3.2
- Kohonen_Const
- Kohonen_Rand_Pat
- PseudoInv
- Randomize_Weights
- Random_Weights_Perc
- RBF_Weights
- RBF_Weights_Kohonen
- RBF_Weights_Redo
- RCC_Weights
- RM_Random_Weights

Niels.Mache@informatik.uni-stuttgart.de

Tue Nov 28 10:30:44 MET 1995