ALEXANDRIA, Va., July 14 -- United States Patent no. 12,682,242, issued on July 14, was assigned to Amazon Technologies Inc. (Seattle).
"Mixing sparsity compression" was invented by Xiaodan Tan (Santa Clara, Calif.), Paul Gilbert Meyer (Jericho, Vt.), Gennady Pekhimenko (Oakville, Canada) and Randy Renfu Huang (Morgan Hill, Calif.).
According to the abstract* released by the U.S. Patent & Trademark Office: "Techniques for compressing a neural network model by mixing compression ratios (sparsity patterns) are described. The weight tensor of a neural network model is divided into weight groups. The pruning cost of compressing the weight values according to a compression ratio is determined for each weight group, and a pruning cost distribut...