Small Subset of Neurons
- CAs should be a small subset of the neurons in the system.
- Some (most, almost all) neurons should participate in multiple
CAs.
- What our small world topology experiments have done is to work
with a small set of orthogonal CAs (neurons are in at most one
CA).
- These could easily be scaled up to a large set of orthogonal
CAs.
- It's not clear how they'd scale up to a large set of
overlapping CAs.
- In the past, we have got hierarchical categories formed
using overlap, but not with appropriate persistence.
