I heard a talk yesterday by Dora Angelaki about multi-sensory integration. Part of the talk was about bayesian inference with multiple noisy sources of information. I know that there has been a lot of theoretical work done on bayesian inference and population coding, but I was wondering if there have been any biological mechanisms proposed for the population activity normalization necessary for these computations?
It depends what you mean by a biological mechanism. If you mean that there should be a protein cascade that implements normalization, that doesn't seem plausible, in my opinion. Normalization in probabilistic population codes is just one of many computations that can be performed in a neural system.
If you're okay with the notion that there's nothing biologically unique about normalization, then Normalization for probabilistic inference with neurons (a paper written by my supervisor, Chris Eliasmith) proposes a spiking neural network performs normalization on probability density functions.