US Patent Issued to Microsoft Technology Licensing on March 17 for "Training a neural network having sparsely-activated sub-networks using regularization" (Washington, California Inventors)

ALEXANDRIA, Va., March 17 -- United States Patent no. 12,579,426, issued on March 17, was assigned to Microsoft Technology Licensing LLC (Redmond, Wash.).

"Training a neural network having sparsely-activated sub-networks using regularization" was invented by Jian Jiao (Bellevue, Wash.), Xiaodong Liu (Redmond, Wash.), Jianfeng Gao (Woodinville, Wash.) and Ruofei Zhang (Mountain View, Calif.).

According to the abstract* released by the U.S. Patent & Trademark Office: "A training technique trains a neural network having sparsely-activated sub-networks. It does so by processing plural batches of training data in two respective passes of the neural network, yielding first prediction information and second prediction information. For each batch...