Fig. 4: Experimental results of pattern classification using a fully perovskite nickelate neural network structure in hardware.
From: Protonic nickelate device networks for spatiotemporal neuromorphic computing
a, Patterns (U, C, S and D) used for this experiment. b, An example of the converted voltage spike trains corresponding to the S pattern. c, The schematic of spatiotemporal processing layer and output layer array connection for a fully NNO-based network. d, Real-time monitored current response from pad 5 and pad 6 with U, C, S and D patterns. e, The voltage applied on the linear output layer proportionally converted by the current output from pad 5 and pad 6. f, Experimental synaptic weight distribution of Pd–Pd spatiotemporal processing array. g, Experimentally measured currents in output layer array with four columns.
