US Patent Issued to GOOGLE on Dec. 30 for "Training robust neural networks via smooth activation functions" (California, Maryland, Washington Inventors)

ALEXANDRIA, Va., Dec. 31 -- United States Patent no. 12,511,545, issued on Dec. 30, was assigned to GOOGLE LLC (Mountain View, Calif.).

"Training robust neural networks via smooth activation functions" was invented by Mingxing Tan (Newark, Calif.), Cihang Xie (Baltimore), Boqing Gong (Bellevue, Wash.) and Quoc V. Le (Sunnyvale, Calif.).

According to the abstract* released by the U.S. Patent & Trademark Office: "Generally, the present disclosure is directed to the training of robust neural network models by using smooth activation functions. Systems and methods according to the present disclosure may generate and/or train neural network models with improved robustness without incurring a substantial accuracy penalty and/or increased comput...