Search International and National Patent Collections

1. (WO2017193769) METHOD AND APPARATUS FOR CONVERSION BETWEEN MACHINE LEARNING MODELS

Pub. No.:    WO/2017/193769    International Application No.:    PCT/CN2017/080844
Publication Date: Fri Nov 17 00:59:59 CET 2017 International Filing Date: Wed Apr 19 01:59:59 CEST 2017
IPC: G06N 99/00
Applicants: TRANSWARP TECHNOLOGY (SHANGHAI) CO., LTD.
星环信息科技(上海)有限公司
Inventors: MEI, Kun
梅鹍
AN, Chunlin
安春霖
CHENG, Dawei
程大伟
LIU, Wanggen
刘汪根
Title: METHOD AND APPARATUS FOR CONVERSION BETWEEN MACHINE LEARNING MODELS
Abstract:
A method and apparatus for conversion between a standalone-mode machine learning model and a distributed machine learning model. Compared to the prior art, the method comprises: acquiring model training parameter information, performing data type conversion or description file parsing on same, and training, according to the model training parameter information, a distributed machine learning model (S13); and converting the distributed machine learning model into a standalone-mode machine learning model (S14). The method further comprises: acquiring a standalone-mode machine learning model (S26); and extracting parameter information of the standalone-mode machine learning model, performing data type conversion or description file parsing on same, and constructing, according to the standalone-mode machine learning model, a distributed machine learning model (S28). The method enables conversion between a standalone-mode machine learning model and a distributed machine learning model, thus greatly enhancing compatibility therebetween.