Abstract: The efficient implementation of inference engines for the DNN (Deep Neural Network) execution requires low power consumption and a small area. However, the activation functions' ...