Learning
- Earlier I talked about dual stable state dynamics. The first
is CA ignition and that's the way the system does short term
memory.
- The second is CA formation which is done by synaptic modification.
A CA is long term memory.
- Synaptic modification is done by a Hebbian learning rule.
- If two neurons co-fire their synaptic strength is increased.
- However, this leaves a lot of latitude, for the actual change.
- It doesn't seem that the neurophysiologists have really figured
out how this works yet.
- We've got a correlatory learning rule which sets the weight
to the frequency that the post-synaptic neuron fires when
the pre-synaptic neuron fires.
- We also use a compensatory mechanism that restricts the
overall synaptic weight of a neuron.
- This is going to remain an open question, and we will
explore new learning rules.
