A Learning Theory for Reward-Modulated Spike-Timing-Dependent Plasticity with Application to Biofeedback | Synapse