The authors built a PyTorch library for processing scalar and vector fields defined on the two-dimensional sphere.
GPU library brings differentiable machine learning to spherical data
The authors implement spherical transforms, convolutions and attention as scalable PyTorch operations designed to preserve the geometry and rotational structure of spherical signals.
Big Tech
Thorsten Kurth · Max Rietmann · Mauro Bisson · Andrea Paris · Alberto Carpentieri · Jean Kossaifi · +3 more
NVIDIA Corporation · California Institute of Technology
Research Digest··2 min read
Kurth and colleagues present torch-harmonics, an open-source library for signal processing and machine learning on spherical domains.
Why this paper
From NVIDIA Corporation and California Institute of Technology
In one line
torch-harmonics provides efficient differentiable spherical harmonic transforms and ML operations on the sphere.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§