U-M Knowledge-Guided Machine Learning (KGML) Workshop: Leading the New Paradigm of AI for Science

What is KGML?

Knowledge-Guided Machine Learning (KGML) is a rapidly emerging field that integrates physics-based principles of scientific knowledge with machine learning algorithms. Unlike traditional machine learning which relies solely on data for learning patterns, KGML incorporates physical laws—often expressed as partial differential equations (PDEs), conservation laws, or symmetries—directly into the model’s design, training, or constraints. As more scientists start to realize the shortfalls of data-driven machine learning, from being uninterpretable, non-extrapolatable, and unaccountable, KGML has emerged as a powerful alternative. By embedding domain knowledge into learning workflows, KGML offers a more reliable, explainable, and scientifically grounded approach to AI. Knowledge-guided machine learning thus becomes the frontier of AI for Science (AI4S).

Why Join Us?

If you are a scientist, engineer, or AI researcher passionate about solving real-world problems with principled and interpretable machine learning, this workshop is for you.

Join us to:

• Learn state-of-the-art KGML tools: Gain hands-on experience with modern frameworks including differentiable simulations, hybrid modeling, and equivariant networks.
• Discover domain-specific strategies: Understand how KGML is already transforming fields from biology and chemistry to environmental science and civil engineering, and how it can transform yours.
• Work on real challenges: Participate in an interdisciplinary hackathon where you will team up with peers from diverse domains to solve scientific problems using KGML.
• Build lasting collaborations: Connect with researchers across institutions to spark new collaborations, exchange ideas, and build a vibrant KGML community.

KGML is not just a method—it’s a mindset. Join us to be part of the next wave of AI for Science!