AccUDNN: A GPU Memory Efficient Accelerator for Training Ultra-deep Deep Neural Networks.
Jinrong GuoWantao LiuWang WangQu LuSonglin HuJizhong HanRuixuan LiPublished in: CoRR (2019)
Keyphrases
- memory efficient
- neural network
- deep architectures
- deep learning
- training process
- parallel implementation
- training algorithm
- pattern recognition
- fuzzy logic
- training set
- feed forward neural networks
- feedforward neural networks
- back propagation
- general purpose
- multiple sequence alignment
- backpropagation algorithm
- parallel processing
- test set
- data structure
- decision trees
- external memory
- iterative deepening
- genetic algorithm