GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism.
Yanping HuangYonglong ChengDehao ChenHyoukJoong LeeJiquan NgiamQuoc V. LeZhifeng ChenPublished in: CoRR (2018)
Keyphrases
- neural network
- training algorithm
- training process
- pattern recognition
- feed forward neural networks
- lightweight
- multi layer
- computationally efficient
- fuzzy logic
- backpropagation algorithm
- multilayer perceptron
- artificial neural networks
- artificial intelligence
- shared memory
- radial basis function network
- error back propagation
- neural network structure
- database
- multi layer perceptron
- online learning
- semi supervised
- data sets