Combining Max Pooling and ReLU Activation Function in Stochastic Computing

In the convolutional neural networks (CNNs) based on stochastic computing (SC), the design of pooling layer is always a key point. Although some SC-based CNNs have been proposed, the circuit output accuracy of pooling layer is not ideal at present. Besides, the pooling layer takes most of the area and power consumption. In this paper, we propose a new kind of SC ReLU with feedback and a combined s...