today announced the development of a high-speed and high-energy-efficiency algorithm and hardware architecture for deep learning processing with less degradations of recognition accuracy.