Abstract
Humans and animals are remarkable at detecting stimuli that predict rewards. While the underlying neural mecha-nisms are unknown, reward influences plasticity of sensory representations in early sensory areas. The underlying changes in excitatory and inhibitory circuitry are however unclear. Recently, experimental findings suggest that the inhibitory circuits can regulate learning. In addition, the inhibitory neurons in superficial layers are highly modulated by diverse long-range inputs, including reward signals. We, therefore, hypothesise that plasticity of in-terneuron circuits plays a major role in adjusting stimulus representations. We investigate how top-down modulation by rewards can interact with local excitatory and inhibitory plasticity to induce long-lasting changes in sensory circuitry. Using a computational model of layer 2/3 primary visual cortex, we demonstrate how interneuron networks can store information about the rewarded stimulus to instruct long-term changes in excitatory connectivity in the absence of further reward. In our model, stimulus-tuned somatostatin-positive interneurons (SSTs) develop strong connections to parvalbumin-positive interneurons (PVs) during reward presentation such that they selectively dis-inhibit the pyramidal layer henceforth. This triggers plasticity in the excitatory neurons, which leads to increased stimulus representation. We make specific testable predictions in terms of the activity of different neuron types. We finally show that this two-stage model allows for translation invariance of the learned representation.