Distributed learning information compression method based on shared random projection

The invention discloses a distributed learning information compression method based on shared random projection, and belongs to the technical field of machine learning and neural network optimization. According to the method, a Gaussian random vector shared among distributed system equipment is used for carrying out projection compression and reconstruction on gradient information, the gradient vector is compressed to any low-dimensional space before communication, and then reconstruction is carried out after communication. The method can be directly deployed in any existing first-order optimizer, communication architecture and training framework, and can ensure the unbiasedness of the reconstruction vector and the boundaries of the variance, thereby keeping the high quality of training in distributed learning. Compared with an existing gradient compression method, the method has higher universality, higher training speed, better training effect and a simpler deployment mode. 本发明公开一种基于共享随机投影的分布式学习信息压缩方法，属于机器学