Metraipolis › archive › dc_892ef0c465be

Functional Degeneracy in Neural Networks: Measurement and Pruning — arXiv:2608.30741

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30741
License (as archived)
CC0-1.0
Archived
2026-10-17T08:39:20.907Z
SHA-256 of the archived text
4b4a960d3fb65880f049b8ece4d84efd2af1c088920f60291e75392a314c31fa
Stable URL of this page
https://metraipolis.com/documento/dc_892ef0c465be

How to cite

Functional Degeneracy in Neural Networks: Measurement and Pruning — arXiv:2608.30741 — arXiv (en), https://arxiv.org/abs/2608.30741 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_892ef0c465be (Metraipolis archive dc_892ef0c465be, sha256 4b4a960d3fb65880f049b8ece4d84efd2af1c088920f60291e75392a314c31fa)
BibTeX
@misc{dc892ef0c465be,
  title = {Functional Degeneracy in Neural Networks: Measurement and Pruning — arXiv:2608.30741},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30741},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_892ef0c465be; sha256 4b4a960d3fb65880f049b8ece4d84efd2af1c088920f60291e75392a314c31fa},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30741. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A central question in modern machine learning is how much a trained model can be compressed without changing its behavior, to reduce the memory, compute and energy required to deploy it. To study this, we quantify functional degeneracy through the behavioral recovery rank, defined as the number of leading behavioral-Hessian eigendirections required to recover a trained model's performance. Using the behavioral recovery rank as a geometric benchmark for compression, we find that structural and magnitude pruning retain more degrees of freedom, even after the task is saturated. This gap suggests that functional redundancy is distributed across parameter directions and is not exposed by individual weights or neurons.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.30741 — 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 2026-10-17T08:13:26.839Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).