Proposes a Frobenius-norm penalty on the structured matrix of a convolutional kernel to bound its singular values around 1 and thereby reduce exploding or vanishing gradients.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Frobenius norm regularization method for convolutional kernels to avoid unstable gradient problem
Proposes a Frobenius-norm penalty on the structured matrix of a convolutional kernel to bound its singular values around 1 and thereby reduce exploding or vanishing gradients.