Title:
【発明の名称】集合型予測装置
Document Type and Number:
Japanese Patent JP3365490
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To provide a device for an aggregated online predictive system having predictive accuracy higher than the performance of a best predictive function among plural prepared predictive functions. SOLUTION: A data space is divided into plural areas and predictive value calculating devices 121-12s are provided corresponding to each of the divided areas. Besides, weight storage devices 131-13s for storing weight information and weight updating devices 141-14s are provided corresponding to each of the predictive value calculating devices. An area discriminating device 11 discriminates the divided area, to which input data belong, and supplies the input data to the correspondent predictive value calculating device and weight updating device. The predictive value calculating device, which receives the input data, finds a predictive value by applying the weight stored in the weight storage device to plural predictive functions and outputs it to a final predictive value calculating device 15. On the basis of the input data and a correct answer value, the weight updating device updates the weight information.
More Like This:
Inventors:
Kenji Yamanishi
Application Number:
JP4818499A
Publication Date:
January 14, 2003
Filing Date:
February 25, 1999
Export Citation:
Assignee:
NEC
International Classes:
G06N99/00; G06F15/18; G06F17/00; G06F19/00; G06N5/04; G06Q10/04; (IPC1-7): G06N1/00; G06F19/00
Domestic Patent References:
| JP9259110A | ||||
| JP635889A | ||||
| JP736847A | ||||
| JP1196132A |
Other References:
井本文雄・他，「雨量情報に基づく数時間先行のダム流入量のオンライン予測法」，電気学会論文誌Ｂ，日本，社団法人電気学会，１９９３年 ３月２０日，Ｖｏｌ．１１３−Ｂ，Ｎｏ．３，ｐｐ．２１３−２２０
寺邊正大、片井修・他，「学習エージェントを用いた協調学習に関する考察」，第３０回人工知能基礎論研究会資料（ＳＩＧ−ＦＡＩ−９７０２），日本，社団法人人工知能学会，１９９７年 ９月２９日，ｐｐ．３７−４２
山西健司，「データ圧縮と学習」，人工知能学会誌，日本，社団法人人工知能学会，１９９７年 ３月 １日，Ｖｏｌ．１２，Ｎｏ．２，ｐｐ．２０４−２１５
山西健司，「拡張型確率的コンプレキシティと情報論的学習理論」，応用数理，日本，１９９８年，Ｖｏｌ．８，Ｎｏ．３，ｐｐ．１４−２９，ＪＳＴ資料番号：Ｌ１１９１Ａ
寺邊正大、片井修・他，「学習エージェントを用いた協調学習に関する考察」，第３０回人工知能基礎論研究会資料（ＳＩＧ−ＦＡＩ−９７０２），日本，社団法人人工知能学会，１９９７年 ９月２９日，ｐｐ．３７−４２
山西健司，「データ圧縮と学習」，人工知能学会誌，日本，社団法人人工知能学会，１９９７年 ３月 １日，Ｖｏｌ．１２，Ｎｏ．２，ｐｐ．２０４−２１５
山西健司，「拡張型確率的コンプレキシティと情報論的学習理論」，応用数理，日本，１９９８年，Ｖｏｌ．８，Ｎｏ．３，ｐｐ．１４−２９，ＪＳＴ資料番号：Ｌ１１９１Ａ
Attorney, Agent or Firm:
Yosuke Goto (1 person outside)