Hybridized learning algorithm based on (Q-learning and A*) algorithms
Other Title(s)
خوارزمية تعلم هجينة مستندة على خوارزميتي (Q-learning and A*)
Dissertant
Thesis advisor
University
Philadelphia University
Faculty
Faculty of Information Technology
Department
Department of Computer Science
University Country
Jordan
Degree
Master
Degree Date
2016
English Abstract
In a line with the current development that were occurred in most of practical life field, such as; artificial intelligence, an increasing need emerges to developing new methods that can improve the motion within agent (e.g.
robot) environment.
The leading process includes two main stages, which are; pathing and learning stages.
Actually, several algorithms and techniques are used during each phase in order to achieve the goal of the leading process, which is to teach the agent how to reach the goal state starting from a specific node (start state) considering several obstacles inside the environment.
Q-learning an A* algorithms are considered common examples for the techniques used during learning and path searching phases respectively.
Actually, when Q-learning algorithm is used in learning, then it may be stuck in a loop of states due to the poor estimation of the environment, this will in turns results in increasing the time needed to reach the goal state.
The aim of this thesis is to model and design a hybridized algorithm that combines both Q-learning and A* algorithm.
The learning process is performed using Q-learning algorithm until it gets stuck in a loop of states, at this point A* algorithm is called and run to provide Q-learning with heuristic values of all four adjacent states.
Q-learning in turns selects the state with the minimum heuristic value as next state and modify Q table according to this value.
The validity of the proposed algorithm was confirmed via comparing it with Q-learning and improved versions of Q-learning that were proposed in the literature
Main Subjects
Information Technology and Computer Science
No. of Pages
49
Table of Contents
Table of contents.
Abstract.
Abstract in Arabic.
Chapter One : Introduction.
Chapter Two : Background and related work.
Chapter Three : Methodology and design.
Chapter Four : Implementation.
Chapter Five : Reults analysis.
Chapter Six : Conclusions and future works.
References.
American Psychological Association (APA)
al-Musalimah, Iyad Radwan. (2016). Hybridized learning algorithm based on (Q-learning and A*) algorithms. (Master's theses Theses and Dissertations Master). Philadelphia University, Jordan
https://search.emarefa.net/detail/BIM-725388
Modern Language Association (MLA)
al-Musalimah, Iyad Radwan. Hybridized learning algorithm based on (Q-learning and A*) algorithms. (Master's theses Theses and Dissertations Master). Philadelphia University. (2016).
https://search.emarefa.net/detail/BIM-725388
American Medical Association (AMA)
al-Musalimah, Iyad Radwan. (2016). Hybridized learning algorithm based on (Q-learning and A*) algorithms. (Master's theses Theses and Dissertations Master). Philadelphia University, Jordan
https://search.emarefa.net/detail/BIM-725388
Language
English
Data Type
Arab Theses
Record ID
BIM-725388