Metraipolis › archive › dc_099ce144eb17
Sparsity-Adaptive Sharpness-Aware Minimization — arXiv:2609.14274
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.14274
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T01:52:54.548Z
- SHA-256 of the archived text
21e0a4cde96b0627dd85c048e2270f6e1c95b7a46bea12e59011481ad94169b7- Stable URL of this page
https://metraipolis.com/documento/dc_099ce144eb17
How to cite
Sparsity-Adaptive Sharpness-Aware Minimization — arXiv:2609.14274 — arXiv (en), https://arxiv.org/abs/2609.14274 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_099ce144eb17 (Metraipolis archive dc_099ce144eb17, sha256 21e0a4cde96b0627dd85c048e2270f6e1c95b7a46bea12e59011481ad94169b7)
BibTeX
@misc{dc099ce144eb17,
title = {Sparsity-Adaptive Sharpness-Aware Minimization — arXiv:2609.14274},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.14274},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_099ce144eb17; sha256 21e0a4cde96b0627dd85c048e2270f6e1c95b7a46bea12e59011481ad94169b7},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.14274. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Deploying deep neural networks in real-world settings requires models that are both compact and robust to common corruptions. However, at deployment-relevant high sparsity, standard pruning pipelines often degrade corruption robustness, and existing sharpness-aware training/pruning approaches provide limited robustness gains. We address this issue by introducing Sparsity-Adaptive Sharpness-Aware Minimization (SA-SAM), which derives a sparsity-dependent SAM/ASAM perturbation radius by keeping the mean absolute perturbation (an $\ell_1$-based proxy) approximately invariant as sparsity increases. As a simple complementary option, we evaluate Magnitude-Weighted Hessian (MWH), derived from a second-order removal-path analysis, yielding an importance proportional to $\mathrm{Diag}(F)_i\,|w_i|$, where $\mathrm{Diag}(F)$ is the diagonal empirical Fisher used as a curvature proxy in our implementation. Across CIFAR-10-C, CIFAR-100-C, and ImageNet-100-C, our approach achieved stronger corruption robustness than the considered pruning baselines at 80--90\% sparsity, while preserving clean accuracy. We additionally quantify the robustness--throughput trade-off by reporting measured inference throughput under sparse execution at deployment-relevant sparsity levels.
Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-15 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.14274 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T01:25:23.564Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-15?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).