Anne Magaly de Paula Canuto
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Anne Magaly de Paula Canuto has a total of 90 co-authors. Here are the top co-authors by number of shared publications:
- João Carlos Xavier Júnior
Universidade Federal do Rio Grande do Norte
11 shared publications - Arthur Costa Gorgonio
Federal University of Rio Grande do Norte (UFRN)
8 shared publications - Cephas Alves da Silveira Barreto
Department of Informatics and Applied Mathematics, Federal University of Rio Grande do Norte, Natal, Brazil
6 shared publications - Karliane Medeiros Ovidio Vale
Federal University of Rio Grande do Norte (UFRN)
5 shared publications - Antonino A. Feitosa Neto
Federal University of Rio Grande do Norte, Natal, Brazil
4 shared publications - Araken de Medeiros Santos
Federal University of Rural do Semi-Árido
4 shared publications
Latest Publications
- Open Access
- 25 August 2026
- Published by Springer Nature in SN Computer Science
- Vol. 7 (7) , 716
Meta-Learning for Selecting Meta-heuristics Based on Iterated Local Search to Solve Permutation Flow-Shop Problem
In the realm of intelligent systems, addressing the NP-hard Permutation Flow-Shop (PFS) problem presents a notable challenge, especially with burgeoning complexities in manufacturing systems. Conventional wisdom relies on metaheuristic techniques to tackle such intricacies, yet the best metaheuristic remains elusive. This paper introduces a new meta-learning framework that leverages pre-trained metaheuristics rooted in Iterated Local Search (ILS) to forecast the most effective metaheuristic for each PFS instance. Notably, our approach pioneers the extraction of novel meta-features from graph representations, discerning the best metaheuristics in 99.7% of cases using a small set of meta-features and baseline classifiers, such as decision trees and support vector machines. Moreover, the interpretability of our meta-model unveils insights into feature impact, showcasing the efficacy of abstract graph computations, such as centrality and connectivity, over bespoke meta-features related to problem size.
- 25 August 2026
- Published by Springer Nature in SN Computer Science
- Vol. 7 (7) , 716
Open AccessMeta-Learning for Selecting Meta-heuristics Based on Iterated Local Search to Solve Permutation Flow-Shop Problem
In the realm of intelligent systems, addressing the NP-hard Permutation Flow-Shop (PFS) problem presents a notable challenge, especially with burgeoning complexities in manufacturing systems. Conventional wisdom relies on metaheuristic techniques to tackle such intricacies, yet the best metaheuristic remains elusive. This paper introduces a new meta-learning framework that leverages pre-trained metaheuristics rooted in Iterated Local Search (ILS) to forecast the most effective metaheuristic for each PFS instance. Notably, our approach pioneers the extraction of novel meta-features from graph representations, discerning the best metaheuristics in 99.7% of cases using a small set of meta-features and baseline classifiers, such as decision trees and support vector machines. Moreover, the interpretability of our meta-model unveils insights into feature impact, showcasing the efficacy of abstract graph computations, such as centrality and connectivity, over bespoke meta-features related to problem size.
- 30 January 2026
- Published by Springer Nature
- p. 377-394
Proposing an Efficient CNN-Based Architecture for Image Processing
- 30 January 2026
- Published by Springer Nature
- p. 377-394
Proposing an Efficient CNN-Based Architecture for Image Processing
- Open Access
- 1 January 2026
- Published by Sociedade Brasileira de Computação in Journal on Interactive Systems
- Vol. 17 (1) , 310-324
Pokémon GO Team Optimization: A Comparative Study of Classic Metaheuristic Algorithms
Pokémon GO team selection can be formulated as a combinatorial optimization problem in which the goal is to generate a three-Pokémon counter-team that maximizes performance against a given rival team under simulated battles. In this study, we establish and evaluate a compact set of classical metaheuristic baselines for this task, namely Simulated Annealing (SA), Tabu Search (TS), Variable Neighborhood Search (VNS), Genetic Algorithm (GA), Ant Colony Optimization (ACO), and Cuckoo Search (CS), using a fixed benchmark of 1,000 rival teams adopted from [da Silva Oliveira et al., 2020]. All methods are assessed under the same dataset, fitness function, and simulator, and we report solution quality (fitness) and elapsed time over repeated runs. The results reveal distinct quality and cost regimes among the evaluated baselines. VNS achieves the highest mean fitness under the adopted stopping conditions, whereas the SA variants provide the lowest runtimes with competitive fitness. We further analyze convergence behavior on the hardest rival teams and characterize the local-search effort of VNS, providing evidence on stabilization patterns and per-iteration workload. These findings deliver reproducible optimization baselines and convergence evidence for Pokémon GO team generation, supporting method selection under different computational budgets and providing reference points for future work on faster convergence and hybrid search strategies.
- 1 January 2026
- Published by Sociedade Brasileira de Computação in Journal on Interactive Systems
- Vol. 17 (1) , 310-324
Open AccessPokémon GO Team Optimization: A Comparative Study of Classic Metaheuristic Algorithms
Pokémon GO team selection can be formulated as a combinatorial optimization problem in which the goal is to generate a three-Pokémon counter-team that maximizes performance against a given rival team under simulated battles. In this study, we establish and evaluate a compact set of classical metaheuristic baselines for this task, namely Simulated Annealing (SA), Tabu Search (TS), Variable Neighborhood Search (VNS), Genetic Algorithm (GA), Ant Colony Optimization (ACO), and Cuckoo Search (CS), using a fixed benchmark of 1,000 rival teams adopted from [da Silva Oliveira et al., 2020]. All methods are assessed under the same dataset, fitness function, and simulator, and we report solution quality (fitness) and elapsed time over repeated runs. The results reveal distinct quality and cost regimes among the evaluated baselines. VNS achieves the highest mean fitness under the adopted stopping conditions, whereas the SA variants provide the lowest runtimes with competitive fitness. We further analyze convergence behavior on the hardest rival teams and characterize the local-search effort of VNS, providing evidence on stabilization patterns and per-iteration workload. These findings deliver reproducible optimization baselines and convergence evidence for Pokémon GO team generation, supporting method selection under different computational budgets and providing reference points for future work on faster convergence and hybrid search strategies.
- 29 September 2025
- Published by Sociedade Brasileira de Computação
- p. 463-474
Analysis of Visual Explainers Using Clustering Techniques
A Inteligência Artificial (IA) tem se tornado cada vez mais presente no cotidiano, ampliando a demanda por métodos que expliquem as decisões tomadas por seus modelos — o que se convencionou chamar de Explainable Artificial Intelligence (XAI). No campo da explicabilidade visual, diversas técnicas e métricas vêm sendo propostas, no entanto, ainda não há consenso sobre como avaliá-las de forma comparativa e confiável. Este trabalho propõe uma nova metodologia para análise de comportamento de explicadores, com base em técnicas de agrupamento (clustering), permitindo avaliar múltiplos explicadores e métricas simultaneamente. Os resultados indicam que o comportamento entre explicadores varia conforme o conjunto de dados e o modelo de explicação utilizado.
- 29 September 2025
- Published by Sociedade Brasileira de Computação
- p. 463-474
Analysis of Visual Explainers Using Clustering Techniques
A Inteligência Artificial (IA) tem se tornado cada vez mais presente no cotidiano, ampliando a demanda por métodos que expliquem as decisões tomadas por seus modelos — o que se convencionou chamar de Explainable Artificial Intelligence (XAI). No campo da explicabilidade visual, diversas técnicas e métricas vêm sendo propostas, no entanto, ainda não há consenso sobre como avaliá-las de forma comparativa e confiável. Este trabalho propõe uma nova metodologia para análise de comportamento de explicadores, com base em técnicas de agrupamento (clustering), permitindo avaliar múltiplos explicadores e métricas simultaneamente. Os resultados indicam que o comportamento entre explicadores varia conforme o conjunto de dados e o modelo de explicação utilizado.
- 30 January 2025
- Published by Springer Nature
- p. 352-367
Dynamicity Analysis in the Selection of Classifier Ensembles Parameters
- 30 January 2025
- Published by Springer Nature
- p. 352-367
Dynamicity Analysis in the Selection of Classifier Ensembles Parameters
- Open Access
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 251-262
Clustering Techniques for Profile Construction Using 5G Access Network Data
A Internet e o desenvolvimento das Tecnologias da Informação e Comunicação expandiram o volume e a diversidade das fontes de dados, abrindo assim novas oportunidades nos setores industriais e acadêmicos à aplicações de tecnologias relacionadas ao Aprendizado de Máquina (AM) e Big Data. Nessa perspectiva, a infraestrutura de Acesso das Redes Móveis vem gerando uma grande quantidade de dados. O presente trabalho tem como objetivo principal a aplicação de algoritmos de agrupamento para a criação de perfis a partir de dados relacionados à indicadores de redes de acesso 5G referentes a tráfego, volume e qualidade de canal. A partir das partições geradas na etapa de agrupamento, espera-se que os perfis encontrados possam servir de base para uma ferramenta de suporte a identificação de falhas, gestão de desempenho e eficiência operacional das redes de acesso.
- 17 November 2024
- Published by Sociedade Brasileira de Computação
- p. 251-262
Open AccessClustering Techniques for Profile Construction Using 5G Access Network Data
A Internet e o desenvolvimento das Tecnologias da Informação e Comunicação expandiram o volume e a diversidade das fontes de dados, abrindo assim novas oportunidades nos setores industriais e acadêmicos à aplicações de tecnologias relacionadas ao Aprendizado de Máquina (AM) e Big Data. Nessa perspectiva, a infraestrutura de Acesso das Redes Móveis vem gerando uma grande quantidade de dados. O presente trabalho tem como objetivo principal a aplicação de algoritmos de agrupamento para a criação de perfis a partir de dados relacionados à indicadores de redes de acesso 5G referentes a tráfego, volume e qualidade de canal. A partir das partições geradas na etapa de agrupamento, espera-se que os perfis encontrados possam servir de base para uma ferramenta de suporte a identificação de falhas, gestão de desempenho e eficiência operacional das redes de acesso.
- 3 May 2024
- Published by Springer Nature in Computational and Applied Mathematics
- Vol. 43 (4) , 1-34
Two classes of ordinal sum implications
- 3 May 2024
- Published by Springer Nature in Computational and Applied Mathematics
- Vol. 43 (4) , 1-34
Two classes of ordinal sum implications
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2000-2005
Two Efficient Training Strategies for DyDaSL: A Dynamic Data Stream Learner Framework with Semi-Supervised Learning
Processing instances quickly and updating models when a concept drift is detected is crucial in the data stream context. However, this process becomes even more challenging in the semi-supervised scenario in which only a few labelled instances are available. A semi-supervised framework called Dynamic Data Stream Learner (DyDaSL) has been proposed in literature to address this issue. This framework generates a fixed size ensemble in the first iteration to classify the remaining data. This paper proposes two extensions for DyDaSL, aiming to optimise the ensemble generation process. In both extensions, the ensemble size starts with only one classifier and it is increased throughout the learning process. In the first extension, a new classifier is included after a stream chunk is presented, while the second extension includes a new classifier when a drift is detected. In order to assess the feasibility of the proposed methods, an empirical analysis is conducted. As a result, these two proposed methods outperformed the standard variation in 128 out of 180 cases (71.11%), showing their effectiveness in all evaluated metrics.
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2000-2005
Two Efficient Training Strategies for DyDaSL: A Dynamic Data Stream Learner Framework with Semi-Supervised Learning
Processing instances quickly and updating models when a concept drift is detected is crucial in the data stream context. However, this process becomes even more challenging in the semi-supervised scenario in which only a few labelled instances are available. A semi-supervised framework called Dynamic Data Stream Learner (DyDaSL) has been proposed in literature to address this issue. This framework generates a fixed size ensemble in the first iteration to classify the remaining data. This paper proposes two extensions for DyDaSL, aiming to optimise the ensemble generation process. In both extensions, the ensemble size starts with only one classifier and it is increased throughout the learning process. In the first extension, a new classifier is included after a stream chunk is presented, while the second extension includes a new classifier when a drift is detected. In order to assess the feasibility of the proposed methods, an empirical analysis is conducted. As a result, these two proposed methods outperformed the standard variation in 128 out of 180 cases (71.11%), showing their effectiveness in all evaluated metrics.
- 12 October 2023
- Published by Springer Nature
- p. 95-109
FlexCon-CE: A Semi-supervised Method with an Ensemble-Based Adaptive Confidence
- 12 October 2023
- Published by Springer Nature
- p. 95-109
FlexCon-CE: A Semi-supervised Method with an Ensemble-Based Adaptive Confidence
- Open Access
- 25 September 2023
- Published by Sociedade Brasileira de Computação
- p. 997-1009
Hyperspectral Imaging: A comparative study of supervised learning algorithms
O presente trabalho mostra um estudo em torno de uma base de dados de imagens hiperespectrais, realizando-se manipulações nesta para a derivação de novas bases de dados para que seja possível observar o comportamento dessas manipulações nos resultados. Além disso, é realizada a automatização do processo de classificação de pixels através da aplicação de algoritmos de aprendizado com diferentes razões de divisão entre as bases de teste e treinamento. Também foram implementados comitês de classificadores como forma de melhorar a acurácia. Os resultados mostram que a Multilayer Perceptron (MLP) é o método com maior acurácia dentre os implementados, atigindo valores superiores a 85% e fornecendo resultados semelhantes aos dos comitês. Dentre as bases analisadas, as que mostraram melhor resultado foi a base original (inalterada) e a que teve seus atributos reduzidos para 20 componentes principais através do algoritmo Principal Component Analysis (PCA). Além disso, devido à consideração dos pixels sem classe definida, obsevou-se uma limitação de acurácia para os algoritmos implementados.
- 25 September 2023
- Published by Sociedade Brasileira de Computação
- p. 997-1009
Open AccessHyperspectral Imaging: A comparative study of supervised learning algorithms
O presente trabalho mostra um estudo em torno de uma base de dados de imagens hiperespectrais, realizando-se manipulações nesta para a derivação de novas bases de dados para que seja possível observar o comportamento dessas manipulações nos resultados. Além disso, é realizada a automatização do processo de classificação de pixels através da aplicação de algoritmos de aprendizado com diferentes razões de divisão entre as bases de teste e treinamento. Também foram implementados comitês de classificadores como forma de melhorar a acurácia. Os resultados mostram que a Multilayer Perceptron (MLP) é o método com maior acurácia dentre os implementados, atigindo valores superiores a 85% e fornecendo resultados semelhantes aos dos comitês. Dentre as bases analisadas, as que mostraram melhor resultado foi a base original (inalterada) e a que teve seus atributos reduzidos para 20 componentes principais através do algoritmo Principal Component Analysis (PCA). Além disso, devido à consideração dos pixels sem classe definida, obsevou-se uma limitação de acurácia para os algoritmos implementados.