GradInit: Learning to Initialize Neural Networks for Stable and Efficient Training.
Chen ZhuRenkun NiZheng XuKezhi KongW. Ronny HuangTom GoldsteinPublished in: NeurIPS (2021)
Keyphrases
- neural network
- training process
- feedforward neural networks
- online learning
- learning algorithm
- supervised learning
- learning machines
- reinforcement learning
- learning stage
- learning process
- active learning
- knowledge acquisition
- neural nets
- serious games
- learning speed
- learning phase
- computer based training
- error back propagation
- multi layer
- learning tasks
- learning systems
- cascade correlation
- connectionist systems
- deep architectures
- minimum error rate
- online training
- training phase
- training set
- learning environment
- machine learning