ALEXANDRIA, Va., June 5 -- United States Patent no. 12,277,502, issued on April 15, was assigned to Microsoft Technology Licensing LLC (Redmond, Wash.).
"Neural network activation compression with non-uniform mantissas" was invented by Daniel Lo (Bothell, Wash.), Amar Phanishayee (Seattle), Eric S. Chung (Woodinville, Wash.) and Yiren Zhao (Cambridge, Great Britain).
According to the abstract* released by the U.S. Patent & Trademark Office: "Apparatus and methods for training a neural network accelerator using quantized precision data formats are disclosed, and in particular for storing activation values from a neural network in a compressed format having lossy or non-uniform mantissas for use during forward and backward propagation train...