ReCA: A Parametric ReLU Composite Activation Function

Abstract

Activation functions have been shown to affect the performance of deep neural networks significantly. While the Rectified Linear Unit (ReLU) remains the dominant choice in practice, the optimal activation function for deep neural networks remains an open research question. In this paper, we propose a novel parametric activation function, ReCA, based on ReLU, which has been shown to outperform all baselines on state-of-the-art datasets using different complex neural network architectures.