Login / Signup
Gradient descent with identity initialization efficiently learns positive definite linear transformations by deep residual networks.
Peter L. Bartlett
David P. Helmbold
Philip M. Long
Published in:
CoRR (2018)
Keyphrases
</>
positive definite
linear transformations
linear transformation
kernel function
covariance matrix
geometric structure
semi parametric
loss function
parameter space
linear model
reproducing kernel hilbert space
optical flow
feature extraction
principal component analysis
model selection