The paper investigates how momentum methods affect large‑batch training in a one‑pass setting using power‑law kernel regression. It derives critical learning rates for SGD, Polyak, and Nesterov, and shows how these rates depend on batch size, momentum, and a capacity exponent. The authors then analyze risk dynamics, optimize final‑step risk under a fixed data budget, and present a three‑regime batch‑size phase diagram that highlights Polyak’s ability to enlarge the critical batch size and Nesterov’s superior data efficiency in the large‑batch regime.
By Jia-Nan Wang, Zixun Huang, Kairui Li, Lei Wu
arXiv:2606. 19179v1 Announce Type: cross Abstract: Stochastic momentum methods such as heavy ball (HB), Nesterov momentum, and variants of Accelerated SGD (ASGD) [Kidambi et al.
By Depen Morwani, Alexandru Meterez, Pranav Nair, Sham Kakade
arXiv:2605. 24316v3 Announce Type: replace Abstract: Mini-batching is central to large-scale optimization, yet its role in statistical scaling laws remains limited.
By Ziyan Chen, Zhongzhu Zhou, Ding-Xuan Zhou
The paper studies momentum-based optimization when updates are sparse, focusing on two tractable models: a least‑squares problem with sparse inputs and a logistic regression problem with a rare class. It derives exact closed‑form second‑moment dynamics and analyzes their high‑dimensional limits across three scaling exponents for sparsity, batch size, and momentum decay. The resulting phase diagram is governed by the ratio of a momentum retention timescale and a learning timescale, revealing regimes where the dynamics match SGD, become unstable, or recover classical heavy‑ball behavior, and showing that oscillatory behavior depends on token sparsity, leading to a spectral conflict for global momentum.
By Katie Everett, Elliot Paquette
arXiv:2609. 40148v1 Announce Type: new Abstract: Power-law learning curves are often treated as fixed properties of a model and its data, although learning-rate and batch-size schedules can change the observed loss.
By Yichen Wang, Fanghui Liu, Yudong Chen
arXiv:2602.06797v3 Announce Type: replace-cross
Abstract: We study optimal learning rate (LR) schedules under the functional scaling law (FSL) framework (Li et al., 2025), which decomposes training d...
By Binghui Li, Zilin Wang, Fengling Chen, Shiyang Zhao, Ruiheng Zheng, Lei Wu
arXiv:2610. 11475v1 Announce Type: new Abstract: Population gradient flow is a common tool for reasoning about how neural networks adapt, including after pretraining.
By Ruoyu Zhao, Mingxuan Zhang, Jianbo Dai, Jiaqi Wu, Chenyu Zhu, Tong Che
The paper presents empirical scaling laws for autoregressive language models, linking prediction loss to model size, data size, and compute, and investigates their theoretical basis using a teacher–student linear RNN framework. In this tractable setting, a stable latent linear RNN generates trajectories while a sketched linear recurrent student is trained via full‑batch WSD gradient descent on next‑token prediction. The study derives explicit approximation, optimization, and statistical scaling laws that depend on the sketch dimension, number of trajectories, and trajectory length, revealing how different power‑law exponents for innovation and initialization covariances affect the rates and crossovers between regimes.
By Ziyan Chen, Zhongzhu Zhou, Peilin Liu, Ding-Xuan Zhou
Population gradient flow is a common tool for reasoning about how neural networks adapt, including after pretraining. We show that it can mispredict finite-batch stochastic gradient descent (SGD) qualitatively, and we trace the discrepancy to a specific mechanism.
arXiv:2607. 12360v1 Announce Type: new Abstract: The cooldown phase of a warmup-stable-decay (WSD) learning-rate schedule, now a default in large-model pretraining, lowers the final training loss in some settings and does nothing in others.
By Subham Singh, Ashutosh Mishra, Subha Raut
arXiv:2605. 24316v2 Announce Type: replace Abstract: Scaling laws provide compact descriptions of how prediction error varies with compute, model size, and data, but existing theory mainly treats single-sample SGD or full data reuse, leaving the role of mini-batching unclear.
By Ziyan Chen, Zhongzhu Zhou, Ding-Xuan Zhou
The paper analyzes the stochastic optimizer Muon, which orthogonalizes an exponential moving average of minibatch gradients, and compares its behavior to the baseline SignSGD on a high‑dimensional matrix‑valued least squares problem. It derives deterministic dynamics showing Muon’s sensitivity to an effective batch size and its failure to resolve small eigenmodes when whitening is exact and the effective batch is small, while Momentum and finite Newton–Shulz iterations can mitigate this. The study presents a phase diagram in terms of data and target exponents, indicating when Muon or SignSGD is preferable based on covariance spectrum steepness and target loss scaling.
By Elliot Paquette, Noah Marshall, Lucas Benigni, Guangyuan Wang, Atish Agarwala, Courtney Paquette