Pith. sign in

REVIEW 11 cited by

On the Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1807.05031 v6 pith:6XQG22ES submitted 2018-07-13 stat.ML cs.LG

On the Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length

classification stat.ML cs.LG
keywords directionslearningrateregionssharpesttrainingalongloss
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Stochastic Gradient Descent (SGD) based training of neural networks with a large learning rate or a small batch-size typically ends in well-generalizing, flat regions of the weight space, as indicated by small eigenvalues of the Hessian of the training loss. However, the curvature along the SGD trajectory is poorly understood. An empirical investigation shows that initially SGD visits increasingly sharp regions, reaching a maximum sharpness determined by both the learning rate and the batch-size of SGD. When studying the SGD dynamics in relation to the sharpest directions in this initial phase, we find that the SGD step is large compared to the curvature and commonly fails to minimize the loss along the sharpest directions. Furthermore, using a reduced learning rate along these directions can improve training speed while leading to both sharper and better generalizing solutions compared to vanilla SGD. In summary, our analysis of the dynamics of SGD in the subspace of the sharpest directions shows that they influence the regions that SGD steers to (where larger learning rate or smaller batch size result in wider regions visited), the overall training speed, and the generalization ability of the final model.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A Defense of the Quadratic Model

    cs.LG 2026-07 conditional novelty 7.0

    Local Taylor-expanded quadratic models reproduce a 150M-parameter LLM's validation loss for up to 10% of training late in the run, and LLM pretraining operates within a factor of 2 of a stochastic or deterministic edg...

  2. Momentum Further Constrains Sharpness at the Edge of Stochastic Stability

    cs.LG 2026-04 unverdicted novelty 7.0

    Momentum SGD exhibits two distinct EoSS regimes for batch sharpness, stabilizing at 2(1-β)/η for small batches and 2(1+β)/η for large batches, aligning with linear stability thresholds.

  3. Characterizing Optimizer-Dependent Training Dynamics Through Hessian Eigenvector Displacement and Localization

    cs.LG 2026-06 unverdicted novelty 6.0

    Hessian eigenvector displacement and inverse participation ratio metrics show SGD stabilizing leading curvature directions while Adam causes more reorganization and parameter localization in MLP training.

  4. Edge of Stability Selectively Shapes Learning Across the Data Distribution

    cs.LG 2026-06 unverdicted novelty 6.0

    Edge of stability acts as a selective mechanism that amplifies learning on data groups with aligned persistent gradients while suppressing others.

  5. Why SGD is not Brownian Motion: A New Perspective on Stochastic Dynamics

    cs.LG 2026-05 unverdicted novelty 6.0

    SGD is reformulated via a master equation from discrete updates, producing a discrete Fokker-Planck equation that predicts non-stationary variance growth proportional to learning rate in flat Hessian directions.

  6. Does Weight Decay Enhance Training Stability?

    cs.LG 2026-05 conditional novelty 6.0

    Weight decay slows progressive sharpening at the edge of stability, inducing damped oscillations in CNNs and a phase transition to sub-2/η sharpness in MLPs driven by parameter-sharpness gradient alignment, yielding m...

  7. Language Models (Mostly) Know What They Know

    cs.CL 2022-07 unverdicted novelty 6.0

    Language models show good calibration when asked to estimate the probability that their own answers are correct, with performance improving as models get larger.

  8. A General Language Assistant as a Laboratory for Alignment

    cs.CL 2021-12 conditional novelty 6.0

    Ranked preference modeling outperforms imitation learning for language model alignment and scales more favorably with model size.

  9. Scaling Laws for Transfer

    cs.LG 2021-02 unverdicted novelty 6.0

    Effective data transferred from pre-training to fine-tuning is described by a power law in model parameter count and fine-tuning dataset size, acting like a multiplier on the fine-tuning data.

  10. Weight-norm Criticality: A Mechanism for Loss Spikes Induced by the Normalization and Weight Decay

    cs.LG 2026-07 conditional novelty 5.0

    Weight decay on scale-invariant weights creates a norm-dependent sharpness boundary; crossing it predicts loss spikes in normalized networks.

  11. Hessian based analysis of SGD for Deep Nets: Dynamics and Generalization

    cs.LG 2019-07 unverdicted novelty 4.0

    Provides Hessian-based theoretical characterizations of SGD dynamics and a scale-invariant generalization bound for deep nets, backed by experiments on synthetic data, MNIST, and CIFAR-10.