Title:
LEARNING DEVICE, LEARNING METHOD, AND LEARNING PROGRAM
Document Type and Number:
WIPO Patent Application WO/2020/044425
Kind Code:
A1
Abstract:
Provided are a learning device, a learning method, and a learning program with which it is possible to obtain an accurately classifying classifier at a low cost. According to an embodiment, a learning device (1) is provided with: an initialization unit (31) which reads data from a training data database (21) that stores labeled data and unlabeled data, and which estimates a class prior probability for the unlabeled data; a data dividing unit (32) which divides the read data into a group (S1) and a group (S2); a weighted risk minimization unit (33) which uses data belonging to the group (S1) to derive a classifier having a low weighted experience risk; a parameter selection unit (34) which uses data belonging to the group (S2) to select hyper-parameters and thereby determine a classifier having a high generalization performance index; and a classifier storage unit (35) which stores the determined classifier.
Inventors:
SAKAI TOMOYA (JP)
LIEW SENG PEI (JP)
LIEW SENG PEI (JP)
Application Number:
PCT/JP2018/031663
Publication Date:
March 05, 2020
Filing Date:
August 28, 2018
Export Citation:
Assignee:
NEC CORP (JP)
International Classes:
G06N99/00
Other References:
SAKAI, TOMOYA ET AL., SEMI-SUPERVISED AUC OPTIMIZATION BASED ON POSITIVE-UNLABELED LEARNING, 16 October 2017 (2017-10-16), pages 1 - 28, Retrieved from the Internet [retrieved on 20181031]
Attorney, Agent or Firm:
IEIRI Takeshi (JP)
Download PDF:
Previous Patent: HYDROGEN DISTRIBUTION PLANNING DEVICE AND HYDROGEN DISTRIBUTION PLANNING METHOD
Next Patent: PHOSPHOR ELEMENT AND ILLUMINATION DEVICE
Next Patent: PHOSPHOR ELEMENT AND ILLUMINATION DEVICE