Blind Thermodynamic Ontology Discovery from Anonymous Experiments

TL;DR

This paper presents a method for discovering thermodynamic ontology from anonymous experiments, allowing machines to learn thermodynamic equations of state. The approach extracts extensive and intensive scaling sectors and recovers their dual cotangent pairing. It enables robust recovery under ill-conditioned mixing.

Why It Matters

The ability to discover thermodynamic ontology is crucial for machine learning models to understand complex systems. This method has practical significance in various fields, including physics and engineering. It can be applied to real-world substances, demonstrating its potential for transferring knowledge across different domains.

Key Findings

• operational identifiability theory
• constructive polynomial-time algorithm
• discrete cyclic concavity