NetReduce: RDMA-Compatible In-Network Reduction for Distributed DNN Training Acceleration.
Shuo LiuQiaoling WangJunyi ZhangQinliang LinYao LiuMeng XuRay C. C. ChuengJianfei HePublished in: CoRR (2020)
Keyphrases
- distributed network
- computer networks
- peer to peer
- training process
- communication overhead
- cooperative
- peer to peer networks
- distributed systems
- local area network
- radial basis function network
- recurrent networks
- multilayer neural network
- network nodes
- camera network
- central server
- network model
- communication cost
- data sets
- training algorithm
- communication networks
- distributed control
- neural network structure
- mobile sensor
- computing platform
- load balance
- network management
- network structure
- training set
- multi agent