On generalization bounds for deep networks based on loss surface implicit regularization.

Masaaki Imaizumi Johannes Schmidt-Hieber

Published in: CoRR (2022)

Keyphrases

generalization bounds
data dependent
empirical risk minimization
learning theory
generalization ability
model selection
kernel machines
ranking algorithm
learning problems
statistical learning theory
linear classifiers
vc dimension
neural network
training data
sample size
uniform convergence