A hardware friendly unsupervised memristive neural network with weight sharing mechanism

•A new hardware friendly unsupervised MNN algorithm is designed with digital integrated circuit, which converts the spike time information of MNN into memristive character directly.•The designed hardware circuits are test by the tasks of image recognition and the recognition accuracy is above 99%.•A weight sharing mechanism is proposed, which can reduce MNNs’ hardware occupancy effectively.•Our network structure and sharing mechanism have a nice resource occupancy trend with the expansion of network scale, which shows the potential to achieve large-scale networks on hardware.