Metraipolis › archive › dc_1e9cc7fd9dc2
NGN: Learning Neural Network Size as a Differentiable Count — arXiv:2609.27291
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.27291
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T10:52:12.954Z
- SHA-256 of the archived text
fe2b4b244927fb8619b7db674a75c92c1a2674221bb4684827ebe6ea7ec651b3- Stable URL of this page
https://metraipolis.com/documento/dc_1e9cc7fd9dc2
How to cite
NGN: Learning Neural Network Size as a Differentiable Count — arXiv:2609.27291 — arXiv (en), https://arxiv.org/abs/2609.27291 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1e9cc7fd9dc2 (Metraipolis archive dc_1e9cc7fd9dc2, sha256 fe2b4b244927fb8619b7db674a75c92c1a2674221bb4684827ebe6ea7ec651b3)
BibTeX
@misc{dc1e9cc7fd9dc2,
title = {NGN: Learning Neural Network Size as a Differentiable Count — arXiv:2609.27291},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.27291},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1e9cc7fd9dc2; sha256 fe2b4b244927fb8619b7db674a75c92c1a2674221bb4684827ebe6ea7ec651b3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27291. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Neural network size is usually chosen before training, separating architecture selection from weight optimization. We introduce the Neurogenesis Network (NGN), a differentiable parameterization for learning how many ordered structural components a model should use. For each ordered component group, one learnable boundary selects an active prefix while the model parameters are trained. The boundary can grow from a compact initialization and can be deployed by discarding components beyond the learned boundary. Controlled experiments examine convergence of the learned boundary, the performance of deployed prefixes, and comparisons with fixed-size models and alternative approaches to learning capacity. We then apply the same mechanism to MLPs, convolutional and graph networks, Transformers, state-space models, LoRA, and adapters. Across these settings, deploying only the learned prefix usually changes performance little, and the selected architectures perform similarly to fixed models trained at the same size. These results show that structural capacity can be optimized directly as a count. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-24 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.27291 — 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 2026-10-08T10:06:16.722Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).