[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Problems, setting parameters for single units

Hi all !

I'm trying to learn on Neural Networks with the PDP++ package.
I've already managed to setup my own little network(9-4-1) and it is learning 
But now I wanted to make some differentiations on how the input-parameters 
are treated (e.g. that I can exclude one or many parameter, without 
redesigning the input-layer and the event-set).
As far as I understand from the documentation, there exist two ways for 
managing this: 
1. Manipulate on the unit object: Creating new child of the Unit-spec and 
setting the activation range to zero (the unit should not get any activation 
any more). Then applying this to a single unit on the input layer.
Result: It is ignored.
2. Creating a new Con-Spec with Uniform weights (mean=0, var=0). This should 
have the advantage, that I can differentiate for which receiver the weight 
should be zero. 
Results: when applied to a single unit, it is ignored
when applied on the Projection the whole layer is treated...
Then I read further in the documentation and found, that ConSpec it always 
used for the whole unit-group and my input-layer was of one group (z=1).
(But why can I change the ConSpec of a single unit ?)
So redesigned the Input-Layer to contain 9 groups each with a single unit 
(z=9) and used the GpFullProjSpec but it didn't help.
I also tried the other Projectionspecs, but none of them let me do what i 
want. Now I'm in a very desperate mood. Can someone please enlight me ?

If this is not the right place for questions like this, can you point me to a 
better place and some Documentation for PDP++ (I have the handbook only).
Mathias Weigt