Overlap and Hierarchy

There wasn't much new in that stuff but I think this stuff is new (and I'm pretty happy with it).
There is no reason that a neuron can't participate in multiple CAs. These are overlapping CAs.
I got some overlapping CAs working and it turns out that the compensatory learning rule helps a lot here.
One of the benefits of this is that categories can share neurons.
An obvious extension was to simulate hierarchies of categories.
So, I developed a hierarchy including Dog, Cat, Rat and Mammal.
I presented Dogs, Cats and Rats. It learned those and also the Mammal category.