AccUDNN: A GPU Memory Efficient Accelerator for Training Ultra-Deep Neural Networks.
Jinrong GuoWantao LiuWang WangChunrong YaoJizhong HanRuixuan LiYijun LuSonglin HuPublished in: ICCD (2019)
Keyphrases
- memory efficient
- neural network
- training process
- parallel implementation
- training algorithm
- feedforward neural networks
- feed forward neural networks
- real time
- external memory
- backpropagation algorithm
- pattern recognition
- iterative deepening
- multi layer perceptron
- back propagation
- multi layer
- multiple sequence alignment
- artificial neural networks
- deep architectures
- error back propagation
- training phase
- genetic algorithm
- feed forward
- training examples
- parallel computing
- neural network training
- training set
- integral image
- deep learning
- neural network model
- general purpose
- supervised learning
- fault diagnosis
- global exponential stability
- search algorithm
- parallel processing
- data structure
- hidden layer