Just kidding, but brain networks are reminiscent of the actor model of computation 
as implemented in programming languages like Erlang.
content
That is, they take in events (data structures associated with an event "type"), combine the information in those events with stored mutable state , and perhaps send out their own events, addressed to specific other neural networks.
Note that the actors/networks are, essentially, singletons . (You don't use a `new` operator to create a new instance and bind it to a name. Networks are statically allocated and not garbage-collected.)
Moreover, the networks "adjacent" to a given network are hardwired. When the brain is "executing" – thinking – it must use predefined constant connections. Connections can be changed by learning, but that's a slow process, not relevant to moment-to-moment thinking.