Memoryless stability-annealed smoothed-sign descent with weighted exponential loss converges in normalized iterates to a Burg-type barrier minimizer on a margin slice, with an explicit S_t^{-1/2} envelope.
Journal of Machine Learning Research , volume =
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
fields
cs.LG 3years
2026 3verdicts
UNVERDICTED 3representative citing papers
A bias-correction framework for stochastic preconditioned optimizers (AdamW, Sophia, Shampoo) using cross-fitted microbatches and delta-method inversion correction yields 0.07-0.15 nat loss reductions on Qwen2.5-0.5B pretraining.
A monograph-length survey claiming deep learning theory can be told as one narrative, from approximation guarantees to emergence.
citing papers explorer
-
Stability Annealing Selects the Implicit Bias of Smoothed Sign Descent: A Rate-Indexed Barrier Path on Separable Data
Memoryless stability-annealed smoothed-sign descent with weighted exponential loss converges in normalized iterates to a Burg-type barrier minimizer on a margin slice, with an explicit S_t^{-1/2} envelope.
-
Correcting Stochastic Update Bias in Preconditioned Language Model Optimizers
A bias-correction framework for stochastic preconditioned optimizers (AdamW, Sophia, Shampoo) using cross-fitted microbatches and delta-method inversion correction yields 0.07-0.15 nat loss reductions on Qwen2.5-0.5B pretraining.
-
From Approximation to Emergence: A Theory of Deep Learning
A monograph-length survey claiming deep learning theory can be told as one narrative, from approximation guarantees to emergence.