Spontaneous Activation
- We've also done some work on Spontaneous Activation
- There's some evidence that neurons fire indepently of
external activation
- This has a useful effect of allowing CAs to grow
- Neurons that aren't in any CAs and are not getting
external activation fire.
- This enables the Hebbian learning rule to strengthen
connections from neurons in CAs to neurons outside it and vice
versa.
- When combined with the compensatory learning rule (below)
this allows these unused neurons to participate in CAs
- Bigger CAs may/should have advantages (future work)
- Spontaneous Activation should keep weights small.
- With the compensatory learning rule it should moderate connection
weights
- From here on out, the connections are biological enough that I'm
going to call them synapses. (Is that reasonable?)
