Title:
MACHINE LEARNING SYSTEM AND METHOD, INTEGRATION SERVER, INFORMATION PROCESSING DEVICE, PROGRAM, AND INFERENCE MODEL GENERATION METHOD
Document Type and Number:
WIPO Patent Application WO/2021/059607
Kind Code:
A1
Abstract:
Provided are: a machine learning system and a machine learning method capable of guaranteeing the accuracy of learning in federated learning; an integration server; an information processing device; a program; and an inference model generation method. According to the present invention, a learning model on each client terminal side and a master model of the integration server are synchronized with each other before each of a plurality of client terminals starts learning. Each client terminal executes machine learning of the learning model by using data stored in a medical institution, and transmits a learning result to the integration server. The integration server divides the client terminals into a plurality of client clusters, integrates learning results for each of the client clusters, and generates master model candidates. The integration server evaluates the inference accuracy of each of the master model candidates, and, upon detection of a master model candidate having an accuracy lower than a threshold, extracts a client terminal which has caused accuracy deterioration from among the client clusters used for generating the master model candidate.
Inventors:
UEHARA DAIKI (JP)
Application Number:
PCT/JP2020/022609
Publication Date:
April 01, 2021
Filing Date:
June 09, 2020
Export Citation:
Assignee:
FUJIFILM CORP (JP)
International Classes:
G06N20/00
Domestic Patent References:
| WO2019022052A1 | 2019-01-31 | |||
| WO2019170785A1 | 2019-09-12 |
Foreign References:
| JP2017174298A | 2017-09-28 |
Other References:
FUNG CLEMENT, YOON CHRIS, BESCHASTNIKH IVAN: "Mitigating Sybils in Federated Learning Poisoning", ARXIV:1808.04866V4, May 2019 (2019-05-01), pages 1 - 16, XP080896917, Retrieved from the Internet [retrieved on 20200827]
Attorney, Agent or Firm:
NAKASHIMA Junko et al. (JP)
Download PDF: