Hybrid-to-NeSy Framework Bridges Mechanistic Models and Neuro-Symbolic AI
The Hybrid-to-NeSy (H2N) framework offers a method for converting hybrid mechanistic and data-driven models into neuro-symbolic (NeSy) AI systems. This strategy reinterprets typical hybrid modeling frameworks as specific instances of a NeSy interface, where mechanistic insights are aligned with the language component, learned elements correspond to the belief aspect, and validity domains along with constraints are situated within the logic segment. H2N produces a defined NeSy inference function and a decomposition of logic and belief, leading to the development of two metrics. This research addresses the challenges of a limited semantic interface for comparing hybrid models across various fields and the oversight of epistemic uncertainty in mechanistic parts. The paper, published on arXiv under ID 2607.22811, details the rationale and methodology behind this translation, aiming to connect process engineering and scientific machine learning with AI.
Key facts
- Hybrid-to-NeSy (H2N) framework translates hybrid mechanistic/data-driven models into neuro-symbolic AI.
- Hybrid models combine first-principles with learned components.
- Common hybrid modeling designs are specified through architectures and training losses.
- H2N places mechanistic knowledge on the language side, learned modules on the belief side, and validity domains with constraints on the logic side.
- H2N yields an explicit NeSy inference functional and a logic-belief decomposition.
- Two metrics are derived from the decomposition.
- The paper is published on arXiv with ID 2607.22811.
- The work bridges process engineering and scientific machine learning with AI.
Entities
Institutions
- arXiv