Mother Tree Optimization for Solving Continuous and Discrete Optimization Problems

Continuous and discrete optimization problems play a signi cant role in di erent academic and industrial disciplines. The main objective of a constraint optimization process is to nd a solution for a problem, that satis es a set of constraints while optimizing a given objective function. The exact or mathematical methods can guarantee the solution if one exists; however, these methods su er from their exponential time cost. Thus, metaheuristic methods have been introduced as e cient approximation approaches for solving real-world optimization problems in a reasonable time frame. In addition, there are always new hard problems that need a robust optimization method to nd the desired solution. This motivates for proposing a robust nature-inspired technique called Mother Tree Optimization (MTO) to solve continuous and combinatorial optimization problems. MTO is inspired by the symbolic relationship between Douglas Fir trees and mycorrhizal fungi network that transfers nutrients between plants of the same and di erent species. In our MTO, the tness values of the feeder (in uencers) candidate solutions are transferred to the non-feeders (in uenced) candidate solutions to improve their solutions. The proposed way of communication between candidate solutions in the MTO algorithm is called Fixed O spring (FO) topology. One issue, when using metaheuristics, is the premature convergence; however, this issue can be handled by nding a good balance between exploitation and exploration processes. In order to asses the performance of MTO, we conduct extensive experiments on a set of known benchmark functions and the results are compared to other known nature inspired techniques. In addition, we have proposed a variant of MTO to tune the weights of a neural network used for a medical diagnostic decision support system, which plays a signi cant role in today's medical technology. Furthermore, we have applied MTO to create a quantum model to solve a hard quantum problem and to overcome the stagnation issue that the other methods su er from. We have also de ned a discrete version of MTO and a discrete variant of the Particle Swarm Optimization (PSO) algorithm, to tackle the following combinatorial optimization problems: Traveling Salesman Problems (TSPs) and Constraints Satisfaction Problems (CSPs). The proposed discrete algorithms are built on a swap operation concept. In order to asses the performance of the proposed methods, we conduct extensive experiments on di erent TSPs instances. In addition, this thesis introduces a new concept called recommendation pool that has been added to MTO and PSO to solve CSPs. In order to asses the performance of the new resulting variants (Mother Tree Optimization for Constraints Satisfaction Problems (DMTO-CSPs) and Mutation Particle Swarm Optimization (MPSO)) we conduct extensive experiments on randomly generated CSPs and di erent real-word problems. DMTO-CSPs and MPSO achieve promising results when solving real-world problems in a reasonable time frame.

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Doctoral -- first

Discipline thesis:degree_discipline

Computer Science

Grantor dc:publisher

Faculty of Graduate Studies and Research, University of Regina

Year dc:date.issued

2021

Author dc:creator

• Korani, Wael Mansour

Advisor dc:contributor.advisor

• Mouhoub, Malek

Committee members dc:contributor.committeemember

• Louafi, Habib
• Butz, Courtney
• Bais, Abdul

Language dc:language.iso

en

DOI

https://doi.org/10.82465/4630

OAI identifier oai:identifier

oai:uregina.scholaris.ca:10294/14940