Adaptive sparse ternary gradient compression for distributed DNN training in edge computing.
Yingchi MaoJun WuXuesong XuLongbao WangPublished in: CCF Trans. High Perform. Comput. (2022)
Keyphrases
- training process
- edge detection
- distributed systems
- distributed environment
- cooperative
- avoid overfitting
- gradient magnitude
- edge information
- multi agent
- image compression
- test set
- multiscale
- orientation histogram
- gradient field
- training phase
- peer to peer
- data compression
- sparse coding
- training examples
- edge map
- compressive sensing
- gradient information
- supervised learning
- active learning
- training set
- edge orientation
- step edges
- image processing