Metraipolis › archive › dc_d8619ddac5b0

Second-Moment Memory in Coordinatewise Adam — arXiv:2608.15824

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.15824
License (as archived)
CC0-1.0
Archived
2026-10-18T19:41:00.276Z
SHA-256 of the archived text
2c276e91a7aec39a3c281a6aebd39d2f9d1fe909a85c947b5f52bd74ba8a86a1
Stable URL of this page
https://metraipolis.com/documento/dc_d8619ddac5b0

How to cite

Second-Moment Memory in Coordinatewise Adam — arXiv:2608.15824 — arXiv (en), https://arxiv.org/abs/2608.15824 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d8619ddac5b0 (Metraipolis archive dc_d8619ddac5b0, sha256 2c276e91a7aec39a3c281a6aebd39d2f9d1fe909a85c947b5f52bd74ba8a86a1)
BibTeX
@misc{dcd8619ddac5b0,
  title = {Second-Moment Memory in Coordinatewise Adam — arXiv:2608.15824},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.15824},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d8619ddac5b0; sha256 2c276e91a7aec39a3c281a6aebd39d2f9d1fe909a85c947b5f52bd74ba8a86a1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15824. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Adam retains a moving average of past squared gradients in its denominator, but the optimization cost of this memory is not well understood. We show that second-moment memory can itself suppress progress toward the optimum even under finite-variance stochastic gradients. For a simple two-point oracle, the expected positive normalized update is $O(M_2^{-1/2})$ after an initialization transient, where $M_2=(1-\beta_2)^{-1}$ is the second-moment memory length. We convert this directional bound, under the stated memory and stepsize scaling, into an average-stationarity lower bound of the same order on a smooth convex problem with normalized gap, smoothness, and variance. Long second-moment memory can slow optimization even when the gradient noise has finite variance.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.15824 — 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 2026-10-18T19:02:44.126Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).