Batch normalization postpones instability in whitened square-loss linear regression by increasing effective learning rate, with explicit no-rising-edge and delayed-onset conditions plus finite self-stabilization.
In: International Conference on Machine Learning, PMLR, pp 247–257 Ahn K, Bubeck S, Chewi S, et al (2023) Learning threshold neurons via edge of stability
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Mechanism Study of Delayed Loss Spikes in Batch-Normalized Linear Models
Batch normalization postpones instability in whitened square-loss linear regression by increasing effective learning rate, with explicit no-rising-edge and delayed-onset conditions plus finite self-stabilization.