Title:
METHOD, NODE AND SYSTEM FOR TRAINING REINFORCEMENT LEARNING MODEL, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2020/062165
Kind Code:
A1
Abstract:
Disclosed are a method, node and system for training a reinforcement learning model, and a storage medium. The training method comprises: a training node acquiring local data, and inputting the local data, as a sample, into a first neural network for training, so as to obtain a first optimal sub-objective function; receiving parameters of a second optimal sub-objective function from a neighbour node; substituting the parameters of the second optimal sub-objective function into the first optimal sub-objective function to obtain the second optimal sub-objective function; and carrying out a weighted average operation on the first optimal sub-objective function and the second optimal sub-objective function, so as to obtain an optimal objective function. By means of the above-mentioned means, the present application can ameliorate the problem of data leakage in a reinforcement learning model training process.
Inventors:
YUAN ZHENNAN (VG)
ZHU PENGXIN (VG)
ZHU PENGXIN (VG)
Application Number:
PCT/CN2018/108766
Publication Date:
April 02, 2020
Filing Date:
September 29, 2018
Export Citation:
Assignee:
BCM SOCIAL CORP (VG)
International Classes:
G06F21/62; G06N99/00; G06Q30/02
Foreign References:
| CN108427891A | 2018-08-21 | |||
| CN107659444A | 2018-02-02 | |||
| CN108520303A | 2018-09-11 |
Attorney, Agent or Firm:
CHINA WISPRO INTELLECTUAL PROPERTY LLP. (CN)
Download PDF: