Get Started
Research questionHow can GPUs efficiently compute unnormalized Gaussian kernel sums with signed weights without custom kernels?Kernel methods repeatedly require unnormalized Gaussian sums, but general-purpose GPU implementations can be slow or memory-intensive. The computation must also support arbitrary signed weights rather than only normalized positive reductions.
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.Fast Gauss Sums via Flash AttentionThe source describes evaluating these sums through flash-attention operations with small input augmentations and no custom GPU code. Evidence covers feature dimensions above 8 in fp16, comparisons with compiled PyTorch and PyKeOps, and linear memory scaling, including speed, memory overhead, and accuracy.research paper · Sep 4, 2026
Related questions
How can LLM-generated GPU kernels remain performant across hardware, inputs, and programming models?How should GPU implementations be compared fairly when their optimization effort is asymmetric?How can sparse balanced signed graph Laplacians be learned efficiently while preserving positive-graph spectral tools?How can multi-shell Leech-lattice 2-bit weights be decoded efficiently for batch-one GPU serving?
Home
Topics
Search
Library