Cognition arises from the coordinated activity of large neuronal populations. In sensory cortex, much of this activity – including the seemingly complex patterns of cortical variability and correlations – can be described to a good approximation by a remarkably small number of global factors. These low-dimensional dynamics must emerge from the interaction of excitatory and inhibitory neuronal populations.
I will first describe how we can quantitatively model the structure of cortical population activity by a surprisingly small number of 'macroscopic variables'. Using large-scale recordings in the primary visual cortex (V1) of anesthetized cat and quietly awake mice, we show that cortical variability is shared across neurons and involves two factors: a multiplicative gain and an additive offset. These two factors determine pairwise correlation and constrain information coding.
I will then talk about our current project on developing a dynamical system model that summarizes the interaction of excitatory and inhibitory populations. To study and manipulate E-I dynamics, we combine electrophysiological recording using multisite silicon probes with dual-wavelength optogenetic to activate pyramidal and parvalbumin-expression neurons independently in V1 of awake mice. A preliminary model is able to capture this dynamics qualitatively and offers possible insights to the underlying mechanism.
Reference: Lin, Xing, & Shapley (2012) Integrate-and-fire vs Poisson models of LGN input to V1 cortex: noisier inputs reduce orientation selectivity. Journal of Computational Neuroscience, 33 559-572