This function call is used to select the different pruning algorithms for neural networks. (See chapter Pruning algorithms). A function call may look like this:
setPruningFunc (function name1, function name2, parameters)
where function name1 is the name of the pruning function and has to be selected from:
MagPruning OptimalBrainSurgeon OptimalBrainDamage Noncontributing_Units Skeletonization
Function name2 is the name of the subordinated learning function and has to be selected out of:
BackpropBatch Quickprop BackpropWeightDecay BackpropMomentum Rprop Std_Backpropagation
Additionally the parameters described below can be entered. If no parameters are entered default values are used by the interpreter. Those values appear in the graphical user interface in the corresponding widget of the pruning window.
Function calls could look like this:
setPruningFunc("OptimalBrainDamage","Std_Backpropagation")
setPruningFunc("MagPruning", "Rprop", 15.0, 3.5, FALSE, 500, 90,
1e6, 1.0)
In the first function call the pruning function and the subordinate learning function is selected. In the second function call almost all parameters are specified. Please note that a function call has to be specified without a carriage return. Long function calls have to be specified within one line. The following text is displayed by the batch interpreter:
Pruning function is now MagPruning
Subordinate learning function is now Rprop
Parameters are: 15.0 3.5 FALSE 500 90 1.0 1e-6 TRUE TRUE
The regular learning function PruningFeedForward has to be set with the function call setLearnFunc(). This is not necessary if PruningFeedForward is already set in the network file.