Title:
OPTIMIZATION DEVICE, OPTIMIZATION METHOD, AND NON-TRANSITORY COMPUTER-READABLE MEDIUM IN WHICH OPTIMIZATION PROGRAM IS STORED
Document Type and Number:
WIPO Patent Application WO/2021/070229
Kind Code:
A1
Abstract:
This optimization device (100) comprises: a setting unit (110) for setting a prescribed non-linear objective function; a policy determination unit (120) for determining, on the basis of the non-linear objective function, the policy to be executed in the on-line optimization of a bandit problem; a policy execution unit (130) for acquiring remuneration as the result of execution of the determined policy; an update rate determination unit (140) for determining, on the basis of the acquired remuneration and the non-linear objective function, the rate of updating of the non-linear objective function by a multiplicative weight update method; and an update unit (150) for updating the non-linear objective function on the basis of the update rate.
More Like This:
Inventors:
ITO SHINJI (JP)
Application Number:
PCT/JP2019/039519
Publication Date:
April 15, 2021
Filing Date:
October 07, 2019
Export Citation:
Assignee:
NEC CORP (JP)
International Classes:
G06N20/00
Other References:
SATYEN KALE, LEV REYZIN, ROBERT E. SCHAPIRE: "Non-Stochastic Bandit Slate Problems", 2010, XP055816594, Retrieved from the Internet [retrieved on 20191122]
Attorney, Agent or Firm:
IEIRI Takeshi (JP)
Download PDF: