• search
    search
  • reviewers
    reviewers
  • feeds
    feeds
  • assignments
    assignments
  • settings
  • logout

The inverse variance-flatness relation in stochastic gradient descent is critical for finding flat minima.

Yu FengYuhai Tu
Published in: Proc. Natl. Acad. Sci. USA (2021)
Keyphrases
  • stochastic gradient descent
  • least squares
  • matrix factorization
  • loss function
  • step size
  • support vector machine
  • collaborative filtering
  • online learning
  • weight vector