Low-power implementation of mitchell's approximate logarithmic multiplication for convolutional neural networks | Publicación