RELIABILITY ASSESSMENT DEVICE AND RELIABILITY ASSESSMENT METHOD

WIPO Patent Application WO/2020/194509

A1

The present invention is provided with: a training data storage unit (102) for storing training data items comprising sets of data items and labels that are information items related to the data items and that are assigned for identifying objects to be identified; a learning unit (103) that, after applying a preset dropout parameter to a neural network model to be learned so as to perform a dropout process thereon, repeats learning for classifying the plurality of labels by using the training data items, the repeated learning being performed until the learning is converged; a model reconfiguration unit (105) for reconfiguring, in accordance with the dropout parameter, a learned model that is the neural network model obtained by having the repeated learning converged and for creating a plurality of different reconfiguration models; an identification unit (107) that identifies the training data items by using the created reconfiguration models and deduces the respective labels of the reconfiguration models; and a classification determination unit (109) that, on the basis of the deduction labels having been deduced, assesses the respective labels of the training data items and classifies the labels of the training data items.

WO/2020/198520	PROCESS AND SYSTEM INCLUDING AN OPTIMIZATION ENGINE WITH EVOLUTIONARY SURROGATE-ASSISTED PRESCRIPTIONS
WO/2022/024167	INFORMATION PROCESSING DEVICE, INFORMATION PROCESSING METHOD, INFORMATION PROCESSING PROGRAM, AND INFORMATION PROCESSING SYSTEM
WO/2018/017423	USING PROXIES TO ENABLE ON-DEVICE MACHINE LEARNING

FUJINO TOMOYA (JP)

PCT/JP2019/012877

October 01, 2020

March 26, 2019

Click for automatic bibliography generation  

MITSUBISHI ELECTRIC CORP (JP)

G06N3/08

JP2018045483A	2018-03-22
JP2015079308A	2015-04-23
JP2018106463A	2018-07-05
JP2018045483A	2018-03-22

GAL, YARIN: "Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning", ARXI V, 4 October 2016 (2016-10-04), XP055467519, Retrieved from the Internet > [retrieved on 20190624]
MAKOTO SATOH, PICTORIAL SYMBOLS, vol. 2016 -CG, no. 3, 1 February 2016 (2016-02-01), pages 1 - 8, XP055744389, ISSN: 2188-8949, Retrieved from the Internet > [retrieved on 20190624]
C. M. BISHOPHIROSHI MOTODA: "Pattern Recognition and Machine Learning; First Volume", ASAKURA PUBLISHING CO., LTD., pages: 174 - 175,450-451
See also references of EP 3926545A4

TAZAWA, Hideaki et al. (JP)

View/Download PDF      PDF Help

Previous Patent: VEHICLE ALARM DEVICE AND VEHICLE ALARM METHOD

Next Patent: ELECTRODE, NON-AQUEOUS ELECTROLYTE BATTERY, AND BATTERY PACK