A new adaptive step-size for SGD, built on the AdaGD Lyapunov function, provably converges in several convex settings without tuned hyper-parameters.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Stochastic Adaptive Gradient Descent Without Descent
A new adaptive step-size for SGD, built on the AdaGD Lyapunov function, provably converges in several convex settings without tuned hyper-parameters.