US 12,073,295 B2
Machine learning model operation management system and method
Souichi Takashige, Tokyo (JP); Daisuke Komaki, Tokyo (JP); and Shin Tezuka, Tokyo (JP)
Assigned to HITACHI, LTD., Tokyo (JP)
Filed by HITACHI, LTD., Tokyo (JP)
Filed on Sep. 23, 2020, as Appl. No. 17/029,812.
Claims priority of application No. 2020-054839 (JP), filed on Mar. 25, 2020.
Prior Publication US 2021/0304070 A1, Sep. 30, 2021
Int. Cl. G06N 20/00 (2019.01); G06F 16/901 (2019.01); G06N 5/04 (2023.01)
CPC G06N 20/00 (2019.01) [G06F 16/9024 (2019.01); G06N 5/04 (2013.01)] 5 Claims
OG exemplary drawing
 
1. A machine learning model operation management system, which is a system for performing process for a task, connected to a task process server outputting task content logs including a plurality of log lines each including plural kinds of data related to task contents, and performing operation management of a machine learning model,
the machine learning model operation management system comprising one or more processors and one or more non-transitory computer-readable storage media, the non-transitory computer-readable storage media having stored thereon at least:
an inference processing server for
performing inference on the basis of the machine learning model, including an identifier of a machine learning model used in inference results of the inference and the inference, and
outputting inference result logs including a plurality of log lines including each of the plural kinds of data related to the inference results;
a model management server for including the plurality of log lines for the inference results; and
a task inference association unit
specifying a task corresponding to the machine learning model based on a feedback pattern definition indicating a correspondence between the learning model used for inference and the task, and
associating the plurality of log lines related to the task contents on the basis of key information which is information associating a specific log line related to the inference results and a specific log line related to the task contents,
wherein the feedback pattern definition is a graph structure that includes a model ID, a task ID, a combination relation, and a combination relation property and represents dependency between components.