Open AccessDesign and Implementation of Course Arrangement Model Based on Reforming Deep Reinforcement Learning
Author(s) -
Lei Zhang,
Fang-qin Xu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1846/1/012052
Subject(s) - reinforcement learning , computer science , curriculum , artificial intelligence , set (abstract data type) , process (computing) , action (physics) , representation (politics) , genetic algorithm , selection (genetic algorithm) , reinforcement , course (navigation) , machine learning , engineering , law , physics , structural engineering , quantum mechanics , politics , political science , programming language , operating system , aerospace engineering
Aiming at the problem of curriculum arrangement, this article abstracts it into two subjects and multiple attributes, and successfully abstracts the problem into a curriculum arrangement model design problem. This paper analyzes the characteristics of the curriculum arrangement problem and combines the characteristics of the deep reinforcement learning algorithm to propose a reformed deep reinforcement learning model. At the same time, the state representation method and action selection method of each stage are designed. Different from general deep reinforcement learning, this article no longer uses action set as the behavior process, and selects the course by optimizing the weight to achieve the purpose of course arrangement. Finally, this paper carried out a comparative experiment to verify the algorithm, which proved that the deep reinforcement learning algorithm based on transformation can effectively solve the problem of curriculum arrangement, and has higher efficiency and higher learning ability than genetic algorithm.