Fig. 6: The neural network structure for VRRAM arrays.
From: Graphene-based 3D XNOR-VRRAM with ternary precision for neuromorphic computing
a A 2-layer MLP neural network used in this work with binary (black/white) input signals. b Schematic and table of the XNOR operation with 1-bit ternary (−1/0/1) synapses. c Weight quantizing to a 1-bit ternary precision scheme, where the sign and three MSB (i.e., condition bits) determine the ternary level (Supplementary Table 3). d The proposed XNOR operation-focused architecture for 3D VRRAM arrays.
