DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification.

Daegun Yoon Sangyoon Oh

Published in: CoRR (2023)

Keyphrases

probabilistic model
parameter estimation
computational model
high level
neural network
edge detection
management system
similarity measure
image processing
data sets
cost function
optical flow
artificial neural networks
mathematical model
experimental data
simulation model
prediction model