MultiAdam: Parameter-wise Scale-invariant Optimizer for Multiscale Training of Physics-informed Neural Networks.
Jiachen YaoChang SuZhongkai HaoSongming LiuHang SuJun ZhuPublished in: CoRR (2023)
Keyphrases
- scale invariant
- multiscale
- neural network
- scale space
- affine invariant
- natural images
- keypoints
- local binary pattern
- scale invariance
- feature detectors
- image representation
- feature transformation
- scale selection
- single scale
- training set
- translation and scale invariant
- wavelet transform
- edge detection
- pattern recognition
- image structure
- affine transformation
- interest point detectors
- training samples
- scale space theory
- scale invariant features
- invariant features
- pairwise
- image analysis