| Home | Trees | Indices | Help |
|
|---|
|
|
object --+
|
base.parameterizedobject.ParameterizedObject --+
|
base.simulation.EventProcessor --+
|
object --+ |
| |
base.sheetcoords.SheetCoordinateSystem --+
|
base.sheet.Sheet --+
|
base.projection.ProjectionSheet --+
|
base.cf.CFSheet --+
|
JointNormalizingCFSheet
A type of CFSheet extended to support joint sum-based normalization.
For L1 normalization, joint normalization means normalizing the sum of (the absolute values of) all weights in a set of corresponding CFs in different Projections, rather than only considering weights in the same CF.
This class makes it possible for a model to use joint normalization, by providing a mechanism for grouping Projections (see _port_match), plus a learn() function that computes the joint sums. Joint normalization also requires having ConnectionField store and return a norm_total for each neuron, and having an OutputFn that will respect this norm_total rather than the strict total of the ConnectionField's weights. At present, CFPOF_DivisiveNormalizeL1 and CFPOF_DivisiveNormalizeL1_opt do use norm_total; others can be extended to do something similar if necessary.
To enable joint normalization, you can declare that all the incoming connections that should be normalized together each have a dest_port of:
dest_port=('Activity','JointNormalize', 'AfferentGroup1'),
Then all those that have this dest_port will be normalized together, as long as an appropriate OutputFn is being used.
|
|||
|
Inherited from |
|||
|
|||
|
Inherited from |
|||
|
|||
|
Inherited from Inherited from Inherited from Inherited from |
|||
|
|||
|
Inherited from Inherited from |
|||
|
|||
Called by the simulation when the EventProcessor is added to the simulation. If an EventProcessor needs to have any code run when it is added to the simulation, the code can be put into this method in the subclass.
|
Apply the weights_output_fn for every group of Projections. The mask is telling which neurons need to be normalized. |
|
| Home | Trees | Indices | Help |
|
|---|
| Generated by Epydoc 3.0beta1 on Thu Feb 21 15:25:08 2008 | http://epydoc.sourceforge.net |