One weird trick for parallelizing convolutional neural networks

作者：Alex Krizhevsky · 发表于：arXiv (Cornell University) · 年份：2014 · DOI：10.48550/arxiv.1404.5997 · 被引用次数：982 · 研究领域：Advanced Neural Network Applications、Stochastic Gradient Optimization Techniques、Neural Networks and Applications

I present a new way to parallelize the training of convolutional neural networks across multiple GPUs. The method scales significantly better than all alternatives when applied to modern convolutional neural networks.