Topology

The topology of a neural network is the way the connections are organised.
This inlcludes uni-directional vs. bi-directional (the brain is uni-directional)
sparsely vs. well connected (the brain is sparsely connected)
neural types, and synaptic weights.
A lot of attractor nets work with well connected, bi-directional nets (Hopfield). It allows a relatively easy use of statistical mechanics. That's great, but not biologically accurate. You also need some other system to move to new states.
Of course, biological neurons change (LTP, STP, and other things).
It's a lot easier to program if you can ignore that (but see learning).
So, if you lay down the topology, you are, in essence, writing code in neurons.