A New Framework for Cortico-Striatal Plasticity: Behavioural Theory Meets In Vitro Data at the Reinforcement-Action Interface

摘要: Operant learning requires that reinforcement signals interact with action representations at a suitable neural interface. Much evidence suggests this occurs when phasic dopamine, acting as prediction error, gates plasticity cortico-striatal synapses, and thereby changes the future likelihood of selecting action(s) coded by striatal neurons. But hypothesis faces serious challenges. First, is inexplicably complex, depending on spike timing, dopamine level, receptor type. Second, there credit assignment problem—action selection occur long before consequent signal. Third, two types output neuron have apparently opposite effects selection. Whether these factors rule out interface how they to produce unknown. We present computational framework addresses first predict expected activity over an operant task for both action-coding neuron, show co-operate promote in compete suppression extinction. Separately, we derive complete model spike-timing dependent from vitro data. then produces predicted necessary extinction task, remarkable convergence bottom-up data-driven top-down behavioural requirements theory. Moreover, complex dependencies are not only sufficient but Validating model, it can account data describing extinction, renewal, reacquisition, replicate experimental plasticity. By bridging levels between single synapse behaviour, our shows striatum acts action-reinforcement