ASYCHRONOUS TRAINING OF MACHINE LEARNING MODEL

专利名称：ASYCHRONOUS TRAINING OF MACHINE

LEARNING MODEL

发明人：Taifeng WANG,Wei CHEN,Tie-Yan LIU,Fei

GAO,Qiwei YE

申请号：US16327679申请日：20170817

公开号：US20190197404A1公开日：20190627

专利附图：

摘要：Various implementations relate to asynchronous training of a machine learningmodel. A server receives feedback data generated by training the machine learning

model from a worker. The feedback data are obtained by the worker with its own trainingdata and are associated with previous values of a set of parameters of the machinelearning model at the worker. The server determines differences between the previousvalues and current values of the set of parameters at the server. The current value mayhave been updated for once or more due to operation of other workers. Then, the servercan update the current values of the set of parameters based on the feedback data andthe differences between values of the set of parameters. Thus, the updating does notonly take the training result of each worker into consideration but also makes propercompensation for delay between different workers.

申请人：MICROSOFT TECHNOLOGY LICENSING, LLC

地址：Redmond WA US

国籍：US

更多信息请下载全文后查看