Zentropy Theory: A Configurational Framework for Entropy and Emergence Across Scales and Domains

2026 10 15 ZENN AI Week ZENN: Toward a Science-Based AI Framework for Emergence Across Scales in Complex Multimodal and Heterogeneous Systems (https://icds.psu.edu/zenn-toward-a-sc...) About Artificial and biological intelligence can both be viewed as open systems that organize information through the exchange of energy under environmental and boundary constraints. Understanding how complex structures, functions and behaviors emerge from heterogeneous components across multiple scales remains a fundamental challenge spanning science and AI. The Zentropy theory provides a rigorous framework for understanding emergence across scales through the integration of quantum mechanics and statistical mechanics. Building upon that foundation, Zentropy-Enhanced Neural Network (ZENN) extends thermodynamic principles to AI and scientific machine learning. ZENN provides a framework for modeling complex multimodal and heterogeneous systems while preserving physical interpretability. The framework has demonstrated strong predictive performance, robustness and generalization in scientific modeling, classification and free-energy landscape reconstruction. ZENN introduces architectrual principles that support interpretability, controllability and stability through configuration-based decomposition and thermodynamic constraints. These features provide a foundation for studying trustworthy and physically grounded AI systems whose behavior can be analyzed across different domains and for different reasons. Learn more by attending this workshop where presentations will be given based on topics like using ZENN for geography, modeling brain states and pollen identification, as well as how to integrate this framework into your own projects. 9:15 a.m. Zi-Kui Liu "Zentropy Theory: Information, Uncertainty, and Emergence Across Scales" 10 a.m. Sarah Ivory "Accelerating paleoecology through automating fossil identification with ZENN: Leveraging intrinsic uncertainty modeling for robust classification" 11 a.m. Baxi Chong "Information theory of locomotion" 1 p.m. Guorong Wu "Turbocharge graph neural network with Zentropy" 2 p.m. Zhenlong Li "Thermodynamic-Inspired Explainable GeoAI: Uncovering Regime-Dependent Mechanisms in Heterogeneous Spatial Systems" 3 p.m. Xianxiao Sun "ZENN-based AI model of Neuroimaging for Autism"