In order to work with various net-models and learning algorithms,
different initialization functions that initialize the components of a
net are required. Backpropagation for example will not work properly
as long as all weights have the same value, since there can be
no gradient computed. Click in the INIT
line of the remote panel to select an init function.
The following initialization functions are available:
All these functions receive their input from the five init parameter fields
in the remote panel. See figure
Here is a short description of the different initialization functions: