Compensatory Learning and Weight Distribution

• It seems pretty clear that most learning is just between the two neurons. (Is it?)
• However, this doesn't mean that you can't use all of the synapses in one or both in considering the change.
• I've done a fair bit of work with compensatory learning.
• The total synaptic weight is distributed across the all of the neurons synapses.
• So, when a neuron fires, the weights of the synapses whose post-synaptic neurons fire go up, but the weights to those that don't fire go down.
• Note that this change can be based on the total synaptic weight entering the post synaptic neuron,
• leaving the pre-synaptic neuron,
• or both.
• There are other ways to balance.