Spike timing dependent plasticity

Ruadhan O'Flanagan, UCSD

The weight-dependence of a spike timing-dependent plasticity learning rule
determines the computational properties of the resulting model neurons.
Neurons sum their synaptic inputs and compare the sum to a threshold, and
emit spikes if the sum exceeds the threshold. If one interprets the
summation as a summation of evidence in favor of a hypothesis, and the
spike as a decision that the total evidence provided by the synaptic
inputs is sufficient to deduce the truth of the hypothesis, then
a specific form of the STDP learning rule can be derived, which
fixes the weight-dependence. The resulting learning rule shows
stable learning and synaptic competition, in addition to admitting
an information-theoretic interpretation. Properties of the
learning rule, and the validity of the interpretation which leads
to it, will be discussed.