Automatic, dynamic, and nearly optimal learning rate specification via local quadratic approximation.
Yingqiu ZhuDanyang HuangYuan GaoRui WuYu ChenBo ZhangHansheng WangPublished in: Neural Networks (2021)
Keyphrases
- learning rate
- learning algorithm
- closed form
- convergence rate
- conjugate gradient algorithm
- optimal solution
- rapid convergence
- worst case
- multilayer neural networks
- adaptive learning rate
- hidden layer
- error function
- convergence speed
- weight vector
- activation function
- dynamic programming
- pairwise
- search space
- objective function