Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training.
Yujun LinSong HanHuizi MaoYu WangBill DallyPublished in: ICLR (Poster) (2018)
Keyphrases
- communication bandwidth
- sensor networks
- image compression
- distributed systems
- data compression
- compression algorithm
- training set
- multi agent
- distributed data
- training samples
- training process
- training algorithm
- peer to peer
- mobile agents
- fault tolerant
- cooperative
- training phase
- computing environments
- distributed network
- deep architectures
- compression ratio
- lightweight
- edge detection
- supervised learning
- data structure