In what sense is STDP optimal ?

Gerstner Wulfram

Laboratoire de Calcul Neuromimétique Brain Mind Institute and I & C EPFL – Building AA-B 1015 Lausanne Switzerland

phone: +41 21 693 67 13 fax : +41 21 693 96 00 email: wulfram.gerstner@epfl.ch

Abstract:

In timing-based neural codes, neurons have to emit action potentials at precise moments in time. We use a modeling framework to show that synaptic connections can automatically adapt their efficacies so that the postsynaptic neuron generates an action potential at a desired firing time defined by the onset of a temporally unprecise fuzzy teaching signal. The optimal strategy of up- and downregulating synaptic efficacies can be described by a two-phase learning window similar to that of Spike-Timing Dependent Plasticity (STDP). Synapses which are activated slightly before the desired firing time are potentiated whereas those that are activated later are depressed. The temporal form of the optimal potentiation window reflects the time course of an excitatory postsynaptic potential. The duration of the optimal depression window is proportional to the duration of the teaching input. Our work suggests that the time course of the STDP function is not as stereotyped as previously thought, but depends upon the task and changes as the postsynaptic neuron improves its temporal precision. Moreover, our theory of optimal STDP gives more importance to the first spike in a postsynaptic burst than to subsequent ones, similar to spike triplet effects previously observed in experiments.