The hyperbolic space is different from the Euclidean space. It has more capacity.
The volume of a ball grows exponentially with its radius!
Think of a binary tree: the number of nodes grows exponentially with depth.
Hyperbolic geometry is better suited to embed data with an underlying hierarchical/heterogeneous structure.
Let’s make deep learning possible in hyperbolic space.
Hyperbolic deep learning sounds fancy, but anybody can understand it and use it.
We provide friendly and intuitive explanations to make it accessible to any data scientist.
Play with the formulas, use the code, make a contribution.