Probabilistic Computation in Spiking Populations

Rich Zemel

As animals interact with their environments, they must constantly update estimates about their states. Bayesian models provide a description of statistically optimal updating based on prior probabilities, a dynamical model, and sensory evidence, and have proved to be consistent with the results of many diverse psychophysical studies. However, little is known about the neural representation and manipulation of such Bayesian information, particularly in populations of spiking neurons. We consider this issue, suggesting an interpretation based on standard neural architecture and activations. We illustrate the approach on a simple spatial navigation example, and then apply it to a sensorimotor integration task that provides a particularly compelling example of dynamic probabilistic computation.