Learning excitatory-inhibitory neuronal assemblies in recurrent networks

Understanding the connectivity observed in brain and how it emerges from local plasticity rules is a grand challenge modern neuroscience. In primary visual cortex (V1) of mice, synapses between excitatory pyramidal neurons inhibitory parvalbumin-expressing (PV) interneurons tend to be stronger for that respond similar stimulus features, although these are not topographically arranged according their preference. The presence such excitatory-inhibitory (E/I) neuronal assemblies indicates stimulus-specific form feedback inhibition. Here, we show activity-dependent synaptic on input output PV generates circuit structure consistent with mouse V1. Computational modeling reveals both forms must act synergy E/I assemblies. Once established, produce competition neurons. Our model suggests can refine circuits actively shape cortical computations.