Learning Rate Perturbation: A Generic Plugin of Learning Rate Schedule towards Flatter Local Minima.
Hengyu LiuQiang FuLun DuTiancheng ZhangGe YuShi HanDongmei ZhangPublished in: CIKM (2022)
Keyphrases
- learning rate
- convergence rate
- learning algorithm
- error function
- hidden layer
- weight vector
- rapid convergence
- convergence speed
- adaptive learning rate
- multilayer neural networks
- training algorithm
- scheduling problem
- activation function
- feature selection
- step size
- reinforcement learning
- machine learning
- convergence theorem