A feedback-feedforward adaptive stepsize law for gradient descent is shown via Lyapunov analysis to achieve O(1/k) last-iterate convergence for convex locally smooth objectives with robustness to inexact gradients.
Provably faster gradient descent via long steps,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.OC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Adaptive control mechanisms in gradient descent algorithms
A feedback-feedforward adaptive stepsize law for gradient descent is shown via Lyapunov analysis to achieve O(1/k) last-iterate convergence for convex locally smooth objectives with robustness to inexact gradients.