Title:
機械学習方法、画像形成装置、および機械学習プログラム
Document Type and Number:
Japanese Patent JP7396078
Kind Code:
B2
Abstract:
To provide a machine learning method, an image forming apparatus, and a machine learning program that can create a learning model that can appropriately perform conveyance control that satisfies productivity in conveyance of media to be conveyed.SOLUTION: A machine learning method is for learning a learning model used for conveyance control of media to be conveyed (sheets 90) by evaluating productivity that is output of the media to be conveyed per unit time from an image forming apparatus 30, and brings the productivity closer to a target value from the positions of the media to be conveyed detected by position detection units (sensors s1 to s6) at the present stage to perform machine learning of outputs from driving sources (motors M1 to M3, clutches C1, C2) to be output next, and thereby performs reinforcement learning of the learning model, and evaluates the productivity. The timing to give a reward is set to a timing before an action value in the reinforcement learning is attenuated.SELECTED DRAWING: Figure 4
Inventors:
Shun Sugai
Koichi Saito
Tomohiro Kiriyama
Koichi Saito
Tomohiro Kiriyama
Application Number:
JP2020012745A
Publication Date:
December 12, 2023
Filing Date:
January 29, 2020
Export Citation:
Assignee:
Konica Minolta, Inc.
International Classes:
G06N20/00; B41J29/38; G06N3/092; H04N1/00
Domestic Patent References:
| JP2019184944A | ||||
| JP2019034836A | ||||
| JP2008181024A |
Attorney, Agent or Firm:
IBC Ichibancho Patent Attorney Corporation