Competition in the temporal domain among neural activities phase-locked to subthreshold oscillations

Abstract.

 In the presence of a subthreshold membrane oscillation, analog information may be encoded in the timing of spike generation phase-locked to the oscillation. With this spike timing neural code, a competitive network of inhibitory spiking neurons was shown to achieve a novel timing mechanism of neural activity selection: the neurons had higher probabilities of becoming winners if they were stimulated earlier in each oscillatory cycle. Here the timing mechanism and its robustness are studied both numerically and analytically, and the conditions to yield a given number of winners (the inhibitory neurons that remain active after the competition) are investigated. The analysis revealed that activity selection with a small number of winners is ensured for broad ranges of values of the parameters such as the strength and time constant of inhibition. In particular, the number of winners is almost unchanged for various timing differences between stimuli to different neurons. This implies that the timing mechanism is useful for such biological information processing as requires perception of a relatively small number of significant stimulus components.