Koo, P.K., Ploenzke, M., 2019. Improving Convolutional Network Interpretability with Exponential Activations.. https://doi.org/10.1101/650804