Fig. 5.

Inference of the pre-synaptic signals with the capacitive neural network. a Schematic of the classifier. A 2 × 3 subarray of the neural network, as highlighted by the red dashed box, was used to classify the two-dimensional input vector consisting of non-overlapping voltage spikes at different rates, R₁ and R₂ spikes per period (see also Supplementary Figure 10). The NPM synapses were initialized to discrete states. b Color maps of the post-synaptic neuron integrate-and-fire rate per input cycle (only the first integrate-and-fire cycle counts) as a function of the spiking rates of the two pre-synaptic inputs