# What's the difference between divisive and soft normalization?

I know that recursive neural integrators (let me know if I need to clarify this term) can be considers soft normalizers, since their feedback loop means that any stimulus eventually saturates the neuron. However, I often see the term "divisive normalization" (which is apparently a synonym for divisive/shunting inhibition confusingly according to Wikipedia and this paper) and I'm not sure what it means and how it relates to soft normalization. They both seem to be regulating the responses of neurons, but they seem to be doing them in different ways. Is that all there is to it?

Explanation of the relation between the two concepts is appreciated.

• Hey, could you maybe provide a definition of soft normalization (ideally, a mathematical one with reference to a paper) ? I only found it here in the documentation of the Nengo library, where it is only vaguely mentioned. Also, I would suggest to add the NEF tag because the terms "soft normalizers" and "recursive neural integrators" seem to only appear in the nef context, as far as I can tell. – awakenting Sep 6 '16 at 14:49