Figure 5.
Software implementation of STDP on discrete weights in spiking neural networks. (A) Temporal evolution of spike pair accumulations a (dimensionless) for causal (black) and anti-causal (gray) spike-timing-dependences. If a crosses the threshold ath (cyan), the weight is updated and a is reset to zero. Pre- and postsynaptic spike trains are generated by a MIP with c = 0.5 and r = 10 Hz. (B) Corresponding weight evolution (solid red) for a 4-bit weight resolution and a LUT configured with n = 30. The weight evolution of the reference synapse model with continuous weights, but a reduced symmetric nearest-neighbor spike pairing scheme is depicted in solid blue. It differs from that of a synapse model with continuous weights and an all-to-all spike pairing scheme (dashed green).