Diagonalwise Refactorization: An Efficient Training Method for Depthwise Convolutions.
Zheng QinZhaoning ZhangDongsheng LiYiming ZhangYuxing PengPublished in: CoRR (2018)
Keyphrases
- computational cost
- clustering method
- high precision
- high accuracy
- significant improvement
- experimental evaluation
- cost function
- optimization algorithm
- computationally efficient
- input data
- dynamic programming
- synthetic data
- objective function
- feed forward neural networks
- error rate
- detection method
- highly efficient
- training process
- feature extraction
- classification accuracy
- multiresolution
- artificial neural networks
- pairwise
- training set
- preprocessing
- computational complexity