Reinforcement learning: Dopamine ramps with fuzzy value estimates

A new study in reinforcement learning theory shows that extending the temporal difference algorithm to unbiased learning under state uncertainty explains the observed ramping behaviour of dopamine neurons.