<ul><li>Machine learning methods for data-driven identification of partial differential equations (PDEs) typically depend on the number of dimensions and coordinates of collected data.</li><li>A new approach has been introduced to make PDE learning coordinate- and dimension-independent, termed as 'spatially liberated' PDE learning.</li><li>The method uses machine learning to predict scalar field systems evolution with exterior calculus formalism, allowing generalization to arbitrary dimensions.</li><li>Numerical experiments on various models show that this approach enables seamless transitions across different spatial contexts.</li></ul>

Towards Coordinate- and Dimension-Agnostic Machine Learning for Partial Differential Equations

Discover more