DEFESA DE PROPOSTA DE TESE
Aluno: Leonardo de Lima Corrêa
Orientador: Prof. Dr. Márcio Dorn
Título: A Memetic Algorithm Framework for Multimodal Continuous Optimization
Linha de Pesquisa: Algoritmos e Otimização
Esta banca ocorrerá excepcionalmente de forma totalmente remota. Interessados em assistir a defesa poderão acessar a sala virtual através do link https://meet.google.com/atx-fuzm-hpu
Prof. Dr. Manuel José Villalobos Cid (Universidad de Santiago de Chile),
Prof. Dr. Conrado Pedebos (University of Southampton)
Prof. Dr. Ricardo Matsumura Araujo (UFPel)
Presidente da Banca: Prof. Dr. Marcio Dorn
Abstract: In spite of the advances regarding computational methods and the wide range of meta-heuristics proposed for multimodal continuous optimization, there is still the need for developing new strategies focused on issues related to the algoritmhs’ performance when applied to hard problems with complicated objective functions. The general idea of this work is centered around the investigation of distinct metaheuristic’s aspects to deal with problems in the multimodal continuous domain. In this sense, we defined the multimodal 3-D protein structure prediction problem as our real case study, which is one of the most important problems in Structural Bioinformatics. Thus, we proposed an adaptive memetic algorithm as a general framework-based method with multiple populations and niching strategies, which incorporates concepts of bio-inspired algorithms for global optimization with separate local improvement. To evaluate the proposed approach, we designed dif-ferent versions of the framework for three scenarios of multimodal optimization: (i) the general framework for single global continuous optimization with multimodal objective function; (ii) the framework with archive strategy for multimodal optimization with more than one global optimum; and (iii) the framework with specific-problem components for the multimodal problem of predicting the 3-D protein structures. With the development of this work, we aimed to create, via a constructive approach, a robust method capable of dealing with the inherent multimodality and issues of a range of optimization functions, while trying to preserve the accurate results. Our focus was also to evaluate the behavior of the presented methods facing different multimodal problems. In general, the memetic algorithm framework was able to perform well on all the aforementioned scenarios of optimization, by reaching promising results when compared with relevant methods re-lated to the corresponding research fields. Nonetheless, despite the obtained results, we highlight that each implemented algorithmic version still needs improvements regarding each one of the delineated case studies to further enhance the method’s performance and results as a whole.
Keywords: Multimodal optimization. metaheuristic. evolutionary algorithm. swarm intelligence. knowledge-based memetic algorithm. structural bioinformatics.