End-to-End Probabilistic Framework for Learning with Hard Constraints

Utkarsh, Danielle C. Maddix, Ruijun Ma, Michael W. Mahoney, Yuyang Wang · 2025

Read End-to-End Probabilistic Framework for Learning with Hard Constraints by Utkarsh, Danielle C. Maddix, Ruijun Ma, Michael W. Mahoney, Yuyang Wang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-07003.txt.

Description

Utkarsh and co-authors published this paper in 2025. It presents ProbHardE2E, a forecasting framework that respects hard constraints: limits a prediction must never break, like physical laws or operating ranges. Many machine learning forecasters ignore limits or patch them later, and they rarely say how unsure they are.

The framework inserts a differentiable projection layer inside the network, so training learns to land inside the allowed region while scoring the forecast honestly. The training score needs no assumptions about the shape of uncertainty, which frees the estimates from bias baked into likelihood-based methods. Nonlinear constraints are supported, widening the range of usable rules. The team demonstrates the system on learning partial differential equations with uncertainty bands and on probabilistic time-series forecasting. A framework handling both, they argue, shows the idea's generality. Applications from energy grids to weather-sensitive logistics could adopt the layer.

Catalog record

CREATORUtkarsh, Danielle C. Maddix, Ruijun Ma, Michael W. Mahoney, Yuyang Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE138 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-07003

Browse the full library · About the Tower of Babel