NVIDIA Corporation received US Patent 12608612B2 on April 21, 2026 for 'Pruning and accelerating neural networks with hierarchical fine-grained structured sparsity.' The patent application (No. 17681967) was filed on February 28, 2022 and contains 15 claims across CPC classifications G06N 3/082, G06N 3/0464, G06N 3/063, G06N 3/084, and G06N 3/09. The technology covers hierarchical structured sparse parameter pruning designed to improve runtime performance and energy efficiency of neural networks by constraining non-zero value distribution according to per-level sparsity degrees at each hierarchy level.
