Modelling Cell Assemblies

by Chris Huyck

How do people think?

One answer is that we use our neurons.

Another answer is that we use areas of our brains for particular activities.

To some extent both of these answers lack rigor. While we can model neurons, we can't see how they would exhibit behavior like we do. Brain areas are too sophisticated for us to model in any real way.

What is needed is some model in between the two.

One that has enough power to model real psychological phenomena, but one that is simple enough to be modelled with current technology.

Cell Assemblies are such a model.

Cell Assemblies, or CAs, consist of reverberating circuits of neurons.

Our model of CAs is complex enough to show completion effects, maintain activation, and learn patterns unsupervised.

CAs should be able to exhibit rule-like behaviour, hierarchies, and choice when given ambiguous input.

The complex recurrent activity is unlike other artificial neural networks.