Accelerating Pooling Layers in Photonic Convolutional Neural Networks
Emilio Paolini, Lorenzo De Marinis, Luca Maggiani, Nicola Andriolli · 2022 27th OptoElectronics and Communications Conference (OECC) and 2022 International Conference on Photonics in Switching and Computing (PSC) · 2022
We propose a photonic neural network where multiply-accumulate and, for the first time, average pooling are realized exploiting a Mach-Zehnder interferometer mesh. Average pooling outperforms the max pooling counterpart, enabling an optical neural network layer.