Reconfigurable spatial-parallel stochastic computing for accelerating sparse convolutional neural networks

The design of image low-pass filter ASIC architecture based on stochastic computing

“…The probability calculation method can make full use of the statistical characteristics of the data, and the calculation of the probability domain has a smaller computational complexity [3].Compared with traditional numerical computing, probabilistic computing is more suitable for hardware implementation, which can construct a new signal processing unit different from traditional computing, and achieve breakthroughs in the design complexity and power consumption cost of information communication systems and modules [4]. Probabilistic computing was first applied to neural networks, and corresponding new hardware architectures [5][6][7] appeared, and then gradually extended to other fields. References [8] proposed an image processing method based on probabilistic calculation method, and references [9][10][11] introduced the relevant application research of probabilistic calculation method in communication signal processing.…”