This paper proves the weak composition-independence conjectures of Kohn, Marchetti, Massarenti, and Mella using deformations of rational curves and a linearized equation kernel. The proof leverages Riemann-Roch theorems to derive a contradiction from tangent fields with controlled poles. By resolving these conjectures, the authors demonstrate that a neural network employing a general polynomial activation function of a sufficiently high degree is identifiable. AI
RANK_REASON The item is an academic paper published on arXiv. [lever_c_demoted from research: ic=1 ai=0.4]
- alphaXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Hugging Face
- Kohn
- Marchetti
- Mella
- Riemann-Roch theorems for differentiable manifolds
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →