… (1) Δ w = A e − ( | t p r e − t p o s t | ) τ t p r e − t p o s t ≤ 0 A > 0 B e − ( | t p r e − t p o s t | )
τ t p r e − t p o s t > 0 B < 0 w is the synaptic weight. A > 0 and B < 0 are usually constant …

… from biological synapses as function of relative timing ΔT = t pos − t pre between the arrival
time t pre of a pre-synaptic spike and the time t pos of generation of a post-synaptic spike. …

… the neuronal dynamics, we consider a simple case where the LIF neuron zi receives weighted
input X(t) = wI(t) from a presynaptic neuron zj (Fig. 1(a)). The rest potential Vrest is set to 0. …

… Here we use Al[t] to denote the residual mapping to be learned as Al[t] = SN(Fl(Sl[… Then
we have Ol[t]=1∧Sl[t] = Sl[t]. Note that using AND may suffer from the same problem as Spiking …

… Indeed, for each image, we added to the threshold of each neuron, V t h r , a random value
drawn from a uniform distribution in range ± α % of V t h r . We evaluated the proposed DCNN …

… -synaptic spike V mem–pre (t) through one of its axons to the … t pos and the pre-synaptic
spike at t pre (see Figure 2). Specifically, as is shown in Figure 3, Δw = ξ(ΔT), with ΔT = t pos − t …

… , to the post- and presynaptic neurons, t i and t j are the corresponding spike times, Δw ij is
… Note that the weight change does not depend on the exact t i − t j value, but only on its sign. …

… over other frameworks, yielding higher speedup when T is larger, eg, up to 11× when T =
32. Second, we tested SNNs in terms of inference with time steps of T = 128, 256, 512, and …

Recently it has been shown that a repeating arbitrary spatiotemporal spike pattern hidden in
equally dense distracter spike trains can be robustly detected and learned by a single …

… We restricted the learning window to [t i −7·τ + ,t i ] for LTP and to [t i ,t i +7·τ − ] for LTD. For
each afferent, we also limited LTP (respectively LTD) to the last (first) presynaptic spike before …