GALA adapts the learning rate through online learning on gradient alignment, achieves a data-adaptive convergence rate for normalized SGD, and shows robust empirical performance across initial learning rates.
High Probability Bounds for a Class of Nonconvex Algorithms with AdaGrad Stepsize
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
-
Online Learning-guided Learning Rate Adaptation via Gradient Alignment
GALA adapts the learning rate through online learning on gradient alignment, achieves a data-adaptive convergence rate for normalized SGD, and shows robust empirical performance across initial learning rates.