Setting β in balanced Adam to achieve a refresh count R_β ≈1000 based on effective learning horizon T_ES improves validation robustness over fixed-β baselines across 11 vision and language experiments.
For each candidate value ofR 0, the rule selects a single beta value throughβ ref = 1−R 0/TES,followed by projection to the closest beta in the grid
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Refresh-Scaling the Memory of Balanced Adam
Setting β in balanced Adam to achieve a refresh count R_β ≈1000 based on effective learning horizon T_ES improves validation robustness over fixed-β baselines across 11 vision and language experiments.