Dynamical analysis of value coding
Information coding in the brain is often viewed in terms of equilibrium levels of activity, but neural circuits exhibit marked temporal dynamics during the decision process. Significant work has focused on the late stage of decision-making, but little is known about the early dynamics of initial value coding. Using a nonlinear differential equation approach, we find that a simple recurrent network model can explain both intial transient dynamics and steady state value coding. This dynamic divisive normalization approach provides a way to link circuit mechanism to circuit function, and we are currently examining decision dynamics with both computational modeling and neurophysiological techniques.