Metraipolis › archive › dc_460c1dfc39cf

Gradient Descent on Two ReLU Neurons: Global Landscape and Bifurcation Dynamics — arXiv:2610.42118

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.42118
License (as archived)
CC0-1.0
Archived
2026-11-02T04:52:34.862Z
SHA-256 of the archived text
38db892a2f944efdbc54f0c8962a4514f4095c327cb6cefa666cc7a5c27d7f79
Stable URL of this page
https://metraipolis.com/documento/dc_460c1dfc39cf

How to cite

Gradient Descent on Two ReLU Neurons: Global Landscape and Bifurcation Dynamics — arXiv:2610.42118 — arXiv (en), https://arxiv.org/abs/2610.42118 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_460c1dfc39cf (Metraipolis archive dc_460c1dfc39cf, sha256 38db892a2f944efdbc54f0c8962a4514f4095c327cb6cefa666cc7a5c27d7f79)
BibTeX
@misc{dc460c1dfc39cf,
  title = {Gradient Descent on Two ReLU Neurons: Global Landscape and Bifurcation Dynamics — arXiv:2610.42118},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.42118},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_460c1dfc39cf; sha256 38db892a2f944efdbc54f0c8962a4514f4095c327cb6cefa666cc7a5c27d7f79},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.42118. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Understanding how gradient descent learns features remains a central challenge in neural network theory. We study this question in a minimal multi-index setting that already exhibits rich multi-phase gradient dynamics: a well-specified two-neuron ReLU teacher-student model under isotropic Gaussian inputs, trained by population gradient flow from small initialization. We show that the dynamics are organized by two directions: the "easy" bisector direction, which carries the leading signal, and the "hard" splitting direction, which governs specialization. To characterize the loss structure and learning behavior, we analyze the population squared-loss landscape and show that every critical point with both student weights nonzero is either a global minimum or a saddle. We then track the gradient-flow trajectory, showing that student neurons first collapse toward a bisector saddle before escaping and specializing to teacher neurons. Our results provide both a landscape and dynamics account of the multi-phase symmetry-breaking behavior that arises in a simple multi-index model and under standard algorithmic and architectural choices.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2610.42118 — 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 2026-11-02T04:45:12.926Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).