A two-part neural network models reward-based training and provides a unified framework in which to study diverse computations that can be compared to electrophysiological recordings from behaving animals.
Teaching signals from "tutor" brain areas should be adapted to the plasticity mechanisms in "student" areas to achieve efficient learning in two-stage systems such as the vocal control circuit of the songbird.
The anticipation of rewards turns out to have its own hedonic value, on top of that of the reward itself; a wide range of behavioral and neurophysiological data suggest that this anticipation is boosted by prediction errors.
Simultaneous 2-photon imaging of striosomes and matrix in mice shows that striosomes preferentially encode reward-predicting cues whereas both striatal compartments demonstrate reward-related activity.
Disruption of right frontopolar cortex with transcranial magnetic stimulation causes selective deficits in exploratory behavior suggesting that different strategies of exploration are implemented by different neural circuits.
While the basal ganglia have long been thought to mediate learning through dopamine-dependent striatal plasticity, their regulation of motor thalamus plays an unexpected and critical role in reinforcement.