Proposes OFF proximal Newton and regularized Newton methods with inexact derivatives that preserve global and local convergence rates for nonconvex composite and unconstrained optimization, plus an adaptive sampling scheme for finite-sum problems.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
OFF (Proximal) Newton-type Methods with Inexact Derivatives for Unconstrained Optimization
Proposes OFF proximal Newton and regularized Newton methods with inexact derivatives that preserve global and local convergence rates for nonconvex composite and unconstrained optimization, plus an adaptive sampling scheme for finite-sum problems.