Move to a Neural Model

The Hopfield net is well connected and the synapses are symmetrical
The brain has say a trillion neurons, and each has around 1000 synapses to and from it, and connections are not symmetrical.
Extensions of the theory of attractor nets show that it is quite robust, but they do differ significantly from our CA model. For instance, due to fatigue it's easy for a CA net to move to a new psuedo-stable state.
So, I think our CA model is not a typical attractor net.
We have discrete steps, neurons are fatiguing leaky integrators, neurons obey Dale's Law, and learning is done by Hebbian learning.
Unlike Hopfield nets, stable states are a series of neurons that tend to be active.
It's not clear how long stable states persist, but in the absence of external stimulation, ours tend to last indefinitely.
Currently, our simulations are about categorising input.
A stimuli is presented, and the net runs to a pseudo-stable state.
Depending on the state that it ends up in, the stimuli is categorised.