New ‘Program-as-Weights’ Paradigm Proposes Coding Fuzzy Functions via Weighted Programs

A new arXiv preprint introduces the “Program-as-Weights” programming paradigm. The approach treats fuzzy functions as programs whose behavior is controlled by weights. Authors argue this differs from traditional neural‑network representations. The paper outlines theoretical foundations for encoding uncertainty in code. Several example applications demonstrate how weighted programs can model fuzzy logic. Potential advantages include clearer interpretability and modular design. The authors discuss limitations and propose directions for future exploration. The work aims to bridge gaps between conventional programming and machine learning.