One weird trick for parallelizing convolutional neural networks
arXiv:1404.5997
Abstract
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.
Cited by in corpus (10)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Deep Image: Scaling up Image Recognition
- Deep Semantic Ranking Based Hashing for Multi-Label Image Retrieval
- Deep Temporal Appearance-Geometry Network for Facial Expression Recognition
- Theano-based Large-Scale Visual Recognition with Multiple GPUs
- Convolutional Neural Networks at Constrained Time Cost
- Multi-path Convolutional Neural Networks for Complex Image Classification
- Web-Scale Training for Face Identification
- Efficient batchwise dropout training using submatrices
- Purine: A bi-graph based deep learning framework