Fig. 4. Acoustic pattern recognition task.

a Design procedure of acoustic pattern (from recording, through transforming, to integrating). b Designed 20 × 20 acoustic pattern, where each pixel has a grayscale value in the range between 0 and 255. c Designed acoustic pattern. d Single-layer artificial neural network (ANN) consisting of input neurons, synapses, and output neurons. e The hardware neural network (HW-NN) comprising the vdW-hybrid synaptic device for the conceptual neural networks. f Weight update method based on operations of “refresh” and “reprogram” for the vdW-hybrid synaptic device. g Acoustic pattern recognition rates based on three ANNs comprising WSe₂, MoS₂, or hybrid synaptic devices, and comparison with the rate based on SW-NN. h Extracted maximum recognition rates and corresponding variations.