What are CAs
- Again, CAs are recurrent networks of neurons.
- There are a lot of neurons in a CA (a number like 30,000 has
been bandied about), and these neurons tend to have lots
of connections to each other and these connections are strong.
- When some of the neurons are activated, they activate the
rest of the neurons.
- This yields a lot of nice properties like:
- The CA can persist longer than a neuron can remain active (~ 5 sec.)
- There are completion effects.
- There is support for connections to a lot of other CAs
(Semantic Nets and XI).
- Neurons can die
- CAs compete allowing Donkey's not to starve.
- Modellers at the neural level include Palm, Hetherington and Shaprio,
Huyck, and Rochester.