Title:
LIGHTENING WEIGHT OF LEARNING MODEL
Document Type and Number:
WIPO Patent Application WO/2023/210546
Kind Code:
A1
Abstract:
In the present invention, an information processing method causes one or more processors included in an information processing device to execute: acquiring of prescribed learning data; performing of machine learning by inputting prescribed data to weighted trained models obtained by assigning respective weights thereto and which include at least two models obtained by lightening the weight of a prescribed learning model using a neural network; acquiring of learning results in a case where machine learning is performed by inputting prescribed learning data to the weighted trained models; performing of supervised learning using learning data including the weighted trained models and the learning results that are obtained through learning using the weighted trained models; and generating of a prediction model for predicting learning results of respective sets of weights in a case where arbitrarily defined learning data is to be inputted by supervised learning.
Inventors:
KUBOTA NOZOMU (JP)
Application Number:
PCT/JP2023/016014
Publication Date:
November 02, 2023
Filing Date:
April 21, 2023
Export Citation:
Assignee:
KUBOTA NOZOMU (JP)
International Classes:
G06N3/082
Domestic Patent References:
| WO2022023022A1 | 2022-02-03 |
Foreign References:
| JP2021521505A | 2021-08-26 | |||
| US20200311552A1 | 2020-10-01 |
Attorney, Agent or Firm:
INABA, Yoshiyuki et al. (JP)
Download PDF: