Title:
MACHINE LEARNING SYSTEM
Document Type and Number:
Japanese Patent JP2022012826
Kind Code:
A
Abstract:
To improve estimation accuracy.SOLUTION: Machine learning is performed using a recurrent neural network in which each data of time series data is composed of a combination of one input value of the same type or a plurality of input values of different type and the time series data is an input value. The time series data is equally divided into a plurality of groups along a time axis, and data of each group corresponding to the same time order viewed from the earliest time to the latest time in each divided group is set as a new data group at every time according to this time order. Machine learning is performed with a new data group being an input value of the recurrent neural network.SELECTED DRAWING: Figure 6
Inventors:
NAKAMURA TOSHIHIRO
Application Number:
JP2020114936A
Publication Date:
January 17, 2022
Filing Date:
July 02, 2020
Export Citation:
Assignee:
TOYOTA MOTOR CORP
International Classes:
G06N3/08; G06N3/04
Attorney, Agent or Firm:
Aoki Atsushi
Shinji Mihashi
Junichi Tsuruta
Koichi Ito
Nobuo Sekine
Tsutomu Kono
Shinji Mihashi
Junichi Tsuruta
Koichi Ito
Nobuo Sekine
Tsutomu Kono
Previous Patent: piston
Next Patent: A method for producing a composite filler, a resin composition containing the same, a polymer compos...
Next Patent: A method for producing a composite filler, a resin composition containing the same, a polymer compos...