US Patent Issued to NVIDIA on Sept. 9 for "Fine-grained per-vector scaling for neural network quantization" (Texas, California Inventors)

ALEXANDRIA, Va., Sept. 10 -- United States Patent no. 12,412,082, issued on Sept. 9, was assigned to NVIDIA Corp. (Santa Clara, Calif.).

"Fine-grained per-vector scaling for neural network quantization" was invented by Brucek Kurdo Khailany (Austin, Texas), Steve Haihang Dai (Union City, Calif.), Rangharajan Venkatesan (San Jose, Calif.) and Haoxing Ren (Austin, Texas).

According to the abstract* released by the U.S. Patent & Trademark Office: "Today neural networks are used to enable autonomous vehicles and improve the quality of speech recognition, real-time language translation, and online search optimizations. However, operation of the neural networks for these applications consumes energy. Quantization of parameters used by the neura...