Principal investigator(s): Sophie Deneve, Christian Machens, David Barrett and Ralph Bourdoukan
Neural networks compute with dynamic sensory and motor variables in a continually changing world. Here we show that networks of integrate-and-fire neurons can implement arbitrary linear dynamical systems by encoding "prediction errors" with their spikes. Our network model is derived from purely functional principles, yet naturally accounts for two puzzling aspects of cortex. First, it provides a rationale for the tight balance and correlations between excitation and inhibition. Second, it predicts asynchronous and irregular firing as a consequence of predictive population coding, even in the limit of vanishing noise. We show that our spiking networks have error-correcting properties that make them far more accurate and robust than comparable rate models. Our approach suggests that spike times do matter when considering how the brain computes, and that the reliability of cortical representations could have been strongly under-estimated.
Boerlin, M. and Denève, S., Spike-Based Population Coding and Working Memory, PLoS Comput. Biol., 7(2), (2011).