torchsom: The Reference PyTorch Library for Self-Organizing Maps (JMLR 2026). GPU-accelerated SOMs with a scikit-learn API, clustering and visualization.
-
Updated
Oct 18, 2026 - Python
torchsom: The Reference PyTorch Library for Self-Organizing Maps (JMLR 2026). GPU-accelerated SOMs with a scikit-learn API, clustering and visualization.
This is a dimensionality reduction project in the course DD2470 Advanced Topics in Visualization and Computer Graphics at KTH Royal Institute of Technology, Stockholm (2024), made by Linnéa Gustafsson.
N-Dimensional alternative to Neural Networks. SLRM-nD uses Simplex Sectoring for neural compression and regression in sparse hyperspaces (1000D+). Zero-latency inference, deterministic weights and ReLU bridge. Zero training, pure geometry.
To associate your repository with the dimensionality-reduction-algorithms topic, visit your repo's landing page and select "manage topics."