A Layer-Based Sparsification Method For Distributed DNN Training.

Yanqing Hu Qing Ye Zhongyu Zhang Jiancheng Lv

Published in: HPCC/DSS/SmartCity/DependSys (2022)

Keyphrases

training process
significant improvement
detection method
optimization algorithm
experimental evaluation

high precision
data sets
distributed systems
optimization method
synthetic data

training samples
high accuracy
support vector machine
preprocessing
input data

least squares
error rate
prior knowledge
objective function