A new research paper explores certificates of positivity and optimality for learning problems where objectives and constraints are not limited to polynomials. The work isolates an axiomatic core of Fischer's constructive Positivstellensätze, proving theorems for abstract function algebras over ordered fields. This framework distinguishes between objective/constraint functions built from continuous or definable operations and the auxiliary primitives used for certificate construction, which must meet specific scalar and closure axioms. The paper provides instances over various function algebras, including those not closed under square roots, and analyzes complexity related to term length and computation graphs. AI
IMPACT This research could advance theoretical underpinnings for complex optimization problems in machine learning.
RANK_REASON The cluster contains an academic paper published on arXiv.
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →