State
- CAs can encode state, and I use them to implement finite
state automata, and thus process.
- So, take part of the neurons, and call them the state neurons.
- Divide them up into (say orthogonal) sets of neurons, say 10 sets.
- Those sets of neurons are CAs.
- It's pretty easy to get simulated CAs to persist indefinitely.
It's kind of hard to get them to persist
for just seconds.
- So turn one of those CAs on, and that's your state.
- If you've got a new input (and an input CA), you can use the
current state and input CAs to transfer to a new state.
- Inhibition can then shut off the original one.
- That's all you need for an FSA, and that enables you to
do pretty sophisticated processing interacting with an environment.
