Hardware
FlashDiffusion: Fused Tiled Kernel Spectral Decomposition
arXiv:2609.38198v1 Announce Type: new Abstract: Diffusion maps, and kernel methods more generally, provide an interpretable nonlinear spectral representation basis for geometric learning. In the geome
arXiv:2609.38198v1 Announce Type: new Abstract: Diffusion maps, and kernel methods more generally, provide an interpretable nonlinear spectral representation basis for geometric learning. In the geometric limit, small bandwidth, these matrices tend to be high rank and thus require materializing dense Gaussian kernels requires O(N^2) memory. We introduce FlashDiffusion, a matrix-free method that evaluates dense Gaussian kernel blocks in fused GPU tiles and couples the eigensolver to an empirical eta-flow that selects the finite-sample resolution scale. A continuation over sample size and bandwidth warm-starts increasingly expensive spectral solves from coarser resolutions.
Related
- TorchKM: A GPU-Oriented Library for Kernel Learning and Model Selection
- Nautilus: An Auto-Scheduling Tensor Compiler for Efficient Tiled GPU Kernels
- Flash-GMM: A Memory-Efficient Kernel for Scalable Soft Clustering
- From Experience to Expertise: Adoption-Aware Memory Learning for Data-Scarce NPU Kernel Synthesis
Source: arXiv cs.LG | 2026-10-01