Title:
LEARNING DEVICE, LEARNING METHOD, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2020/161935
Kind Code:
A1
Abstract:
[Problem] To enable an appropriate temperature to be determined automatically in deep learning using distillation. [Solution] This learning device is equipped with: a first estimation unit that uses a temperature parameter to carry out estimation on the basis of a student model; a second estimation unit that uses the temperature parameter to carry out estimation on the basis of a teacher model; and a temperature calculation unit that calculates the temperature parameter on the basis of estimation information generated by the first estimation unit and the second estimation unit.
Inventors:
ISHII ASUKA (JP)
Application Number:
PCT/JP2019/026672
Publication Date:
August 13, 2020
Filing Date:
July 04, 2019
Export Citation:
Assignee:
NEC CORP (JP)
International Classes:
G06N3/08
Domestic Patent References:
| WO2018051841A1 | 2018-03-22 |
Foreign References:
| US20180268292A1 | 2018-09-20 |
Other References:
MOSNER, LADISLAV ET AL.: "IMPROVING NOISE ROBUSTNESS OF AUTOMATIC SPEECH RECOGNITION VIA PARALLEL DATA AND TEACHER-STUDENT LEARNING", ARXIV, 11 January 2019 (2019-01-11), pages 1 - 5, XP033565972, Retrieved from the Internet [retrieved on 20190423]
Attorney, Agent or Firm:
NAKAMURA, Toshinobu et al. (JP)
Download PDF:
Previous Patent: COMPOSITE, METHOD FOR PRODUCING SAME, AND COATED METAL SHEET
Next Patent: EYELID SPECULUM
Next Patent: EYELID SPECULUM