Learning by the dendritic prediction of somatic spiking.

Abstract in brief

• Recent research shows that spike-timing-dependent plasticity (STDP) is influenced not only by the timing of pre- and postsynaptic neurons but also by local dendritic potentials, which act as a third factor in plasticity.
• The authors propose a simple neuron model with a biologically plausible learning rule for dendritic synapses that aims to minimize the difference between the neuron's overall firing and the local dendritic potential, accounting for stochastic noise and direct soma input.
• Depending on the type of input, this learning rule can support supervised or unsupervised learning, and when combined with reward signals, it can also facilitate reinforcement learning, making it versatile for different learning contexts.]