Emil Wärnberg

This paper proposes a model where vector-valued dopamine signals enhance learning of continuous outputs in the striatum using a recurrent network and a local plasticity rule.

The paper shows that all three major striatal output pathways represent action space and value similarly during motor and decision tasks.

The paper maps the molecular and spatial organization of striatal projection neurons, identifying new subtypes and spatial domains in both mice and non-human primates.

This paper demonstrates that a biologically plausible spiking neuronal network can exhibit low-dimensional activity, showing that learning is constrained by the intrinsic manifold of network activity.