Benefits of Overparameterized Convolutional Residual Networks: Function Approximation under Smoothness Constraint.
Hao LiuMinshuo ChenSiawpeng ErWenjing LiaoTong ZhangTuo ZhaoPublished in: ICML (2022)
Keyphrases
- function approximation
- smoothness constraint
- reinforcement learning
- learning tasks
- temporal difference learning algorithms
- radial basis function
- optical flow
- motion segmentation
- spatially varying
- energy function
- data mining
- image intensity
- higher order
- image classification
- linear combination
- image sequences
- function approximators
- connectionist networks
- machine learning