AlexNet
lastmod 2026-11-10created 2016-01-262 edits
AlexNet is a convolutional neural network (CNN) that beat existing approaches by a wide margin at the 2012 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) and set off the deep learning boom. It consists of five convolutional layers and three fully connected layers, and is notable for introducing the ReLU activation function, dropout, data augmentation, and parallel training on two GPUs. It became the starting point for later deep CNN designs such as VGG, GoogLeNet, and ResNet.