A Top-Down Approach to SNN-STDP Networks

This chapter presents a top-down optimization-based theory describing spiking cortical ensembles equipped with Spike-Timing-Dependent Plasticity (STDP) learning, as empirically observed in the visual cortex (as opposed to the bottom-up SNN-STDP setup presented in most prior works). In contrast to empirical parameter search used in most previous works, this chapter also provides novel theoretical grounds for SNN and STDP parameter tuning which considerably reduces design time. Using this generic framework, a class of global and action-based feature descriptors is built for event-based cameras and assessed on the N-MNIST and the IBM DVS128 Gesture datasets. Significant accuracy improvements are reported compared to state-of-the-art STDP-based systems (+9.3% on N-MNIST, +7.74% on IBM DVS128 Gesture).