The Disharmony Between BN and ReLU Causes Gradient Explosion, but is Offset by the Correlation Between Activations.

Inyoung Paik Jaesik Choi

Published in: CoRR (2023)

Keyphrases

bayesian networks
neural network
correlation coefficient
image processing
artificial intelligence
machine learning
case study
medical images
database
belief networks
parameter learning
image gradient
piecewise constant
bayesian belief networks
gradient information
gradient method
gradient direction