Bio-inspired temporal-decoding network topologies for the accurate recognition of spike patterns

In this paper will be presented simple and effective temporal-decoding network topologies, based on a neuron model similar to the classic Leaky Integrate-and-Fire, but including the spike latency effect, a neuron property able to take into account that the firing of a given neuron is not instantaneous, but it occurs after a continuous-time delay depending on the inner state. These structures are able to detect spike sequences composed of pulses belonging to neuron ensembles, exploiting basic biological neuron mechanisms. According to the biological counterpart, with these structures is possible to achieve a high temporal accuracy, but also deal with the natural variability present in spike trains. In addition, the connection of these neural structures at a higher level make possible to afford some pattern recognition problems, operating a distributed and parallel input data processing.