Production of Observables. The behavior of individual neurons 26 3.3.9 Representation of spiking messages 27
Production of Observables. The behavior of individual neurons. The DPSNN-STDP code can produce files tracing several observables (list of individual spiking times and spiking neuron identity, mean spiking rates, membrane potentials, synaptic values). The code is equipped with facilities for distributed measurement of time spent on execution of individual routines and sections of the code based on the MPI_WTIME function. Figure 3-3 A sample trace of 320 ms of spiking activity, produced by the DPSNN-STDP code. In this case, it is the simulation of a single neural column, composed by 1000 neurons (80% excitatory RS, 20% inhibitory FS Izhikevich neurons). Above, each dot in the rastergram represents a spiking event. Below, the traces of the membrane potential of two excitatory neurons. If necessary, the trace of the evolution of the membrane potential and other state variables of individual neurons can be activated. The membrane potential of a “resting” neuron fluctuates around a -70mV potential as a result of its own activities and of the perturbation produced by signals produced by other neurons. When the neuron decides to “fire”, its membrane potential (v state variable) starts to climb to positive voltages. If a “spike” happened, the Izhikevich rules drops the voltage down to its after spike potential, but the internal variable u keeps a memory of the past. This event is the “spike” of an individual neuron, which is propagated by the axon, and reaches a set of synapses after a time delay, specific for each synapse, which depends on the distance travelled. When reached by the spike, each synapse produces a perturbation of the membrane potential of the target, which depend on its “strength” (here, we consider the simplistic case of the neural soma and dendritic arborization represented by a single “compartment”, i.e. a single u(t)-v(t) pair in the case of the Izhikevich equation). The picture reports about 320 ms of collective spiking activity of a single column of 1000 neurons (800 RS excitatory, 200 FS inhibitory), and the evolution of the individual membrane potential of two neurons. In the “rastergram” the horizontal axis is the simulation time, the vertical axis the identifier of individual neurons. Each dot in the rastergram represents a spiking event.
