← Library · Advanced concept

Geometric Deep Learning

Geometric deep learning extends deep learning techniques to handle data structured on non-Euclidean spaces, such as graphs, manifolds, and 3D shapes. This involves developing neural network architectures that respect the intrinsic geometry and symmetries of the data. For example, graph neural networks (GNNs) are a prominent type of geometric deep learning model designed for data represented as graphs.

In plain terms

It's like transitioning from drawing on a flat paper (Euclidean data) to drawing on a crumpled ball or a complex sculpture (non-Euclidean data), requiring new tools that understand the curves and connections.

Why it matters

It unlocks the power of deep learning for a vast array of real-world data structured as graphs (social networks, molecules) or manifolds, previously difficult for traditional deep learning models.

Learn one new AI thing every day.

Daily Deck sends you seven plain-English cards like this every morning. Free.

Start free