Compensatory Learning Rule
- A CA for a sparse pattern should be learnable; so should
a CA for a dense pattern
- A dense pattern may allow enough weight (from associations) to
allow spontaneous activation to cause a CA to develop
- However, if neurons are only stimulated by a weak pattern,
a CA should be allowed to form.
- The Compensatory Learning Rule allows a wider range of patterns
to form
- The Compensatory Learning Rule encourages the synaptic weights
of a neuron toward a moderate value
- If the total synaptic weight is below the ideal value,
strengthening is increased, and weakening is decreased.
- If the total synaptic weight is above the ideal value,
strengthening is decreased, and weakening is increased.
- Potential CAs associated with sparse patterns are encouraged.
- Dense patterns are weakened.
- It's still Hebbian, so I think it's fair
- It may also prevent Simulated Epilepsy
- Does anyone know of any physiological evidence for this?
