Geometric Intelligence:
From Vision to Scientific Discovery
Vision models excel at capturing spatial structures, scientific applications demand a higher standard: models must be reliable, interpretable, and strictly bound by physical laws rather than just visually plausible. This workshop positions geometry as the fundamental bridge between these two worlds, exploring how spatial inductive biases, symmetry, and structured representations can empower machine learning models to look beyond the surface, moving from simply seeing the world to reasoning, generalizing, and uncovering the hidden structural principles of the natural sciences.
Our topics include but are not limited to:
- Geometric deep learning and equivariance
- Geometry-aware 3D vision and neural scene representations
- Physics-aware and simulation-informed models
- Geometric structures for scientific discovery
- Foundation models for geometric and physical reasoning
- Symmetry and invariance in visual perception
- Topological data analysis and deep learning
- Structured representations for 3D scene understanding
- Applications in biology, robotics, and materials science
