Open Journal of Optimization, 2017, 6, 65-84 http://www.scirp.org/journal/ojop ISSN Online: 2325-7091 ISSN Print: 2325-7105 A Novel Approach Based on Reinforcement Learning for Finding Global Optimum Cenk Ozan 1 , Ozgur Baskan 2 , Soner Haldenbilen 2 1 Department of Civil Engineering, Faculty of Engineering, Adnan Menderes University, Aydin, Turkey 2 Department of Civil Engineering, Faculty of Engineering, Pamukkale University, Denizli, Turkey Abstract A novel approach to optimizing any given mathematical function, called the MOdified REinforcement Learning Algorithm (MORELA), is proposed. Al- though Reinforcement Learning (RL) is primarily developed for solving Mar- kov decision problems, it can be used with some improvements to optimize mathematical functions. At the core of MORELA, a sub-environment is gen- erated around the best solution found in the feasible solution space and com- pared with the original environment. Thus, MORELA makes it possible to discover global optimum for a mathematical function because it is sought around the best solution achieved in the previous learning episode using the sub-environment. The performance of MORELA has been tested with the re- sults obtained from other optimization methods described in the literature. Results exposed that MORELA improved the performance of RL and per- formed better than many of the optimization methods to which it was com- pared in terms of the robustness measures adopted. Keywords Reinforcement Learning, Mathematical Function, Global Optimum, Sub-Environment, Robustness Measures 1. Introduction If ( ) f x is a function of decision variables, where x S ∈ , S is the feasible search space and n S R ⊆ , an optimization problem can be defined as finding the value of best x in S that makes ( ) f x optimal for all x values. Despite the fact that different meta-heuristic algorithms have been improved especially in last two decades, the contributions of Reinforcement Learning (RL) to this area are still limited comparing to others. Numerous studies such as genetic algo- rithm based methods [1] [2], ant colony based algorithms [3] [4], harmony How to cite this paper: Ozan, C., Baskan, O. and Haldenbilen, S. (2017) A Novel Ap- proach Based on Reinforcement Learning for Finding Global Optimum. Open Journal of Optimization, 6, 65-84. https://doi.org/10.4236/ojop.2017.62006 Received: March 31, 2017 Accepted: June 25, 2017 Published: June 28, 2017 Copyright © 2017 by authors and Scientific Research Publishing Inc. This work is licensed under the Creative Commons Attribution International License (CC BY 4.0). http://creativecommons.org/licenses/by/4.0/ Open Access DOI: 10.4236/ojop.2017.62006 June 28, 2017