Events
Divulgação - Defesa Nº 254
Aluno: João Victor Tinoco de Souza Abreu
Título: “Extracting Interpretable Classification Models via Readability-Enhanced Genetic Programming”
Orientador: Fernando Buarque de Lima Neto - (PPGEC)
Co-orientador: Denis Mayr Lima Martins
Examinador Externo: Diego Marconi P. Ferreira Silva - (UNICAP)
Examinador Interno: Cleyton Mário de Oliveira Rodrigues - (PPGEC)
Data-hora: 04 de Agosto de 2022 às 14:30h.
Local: Formato Remoto (https://meet.google.com/pro-eama-hjq)
Resumo:
As the impact of Machine Learning (ML) on business and society grows, there is a need for making ML- based decisions transparent and interpretable, especially in the light of fairness, and to avoid bias, and discrimination. It is known that the high-level applications of complex scenarios require more powerful models, such as Deep Learning (DL) models. Since the user needs to understand the functional details of those models, that is, how the model’s produce their outcomes. This research aims at helping on that front. Even though the use of opaque ML models (OM) for decision-making support trends in many application fields, little is known on revealing how the iteration with the user is valuable and what features and parameters should be used to clarify such OMs. This need for transparency motivated this research. Moreover, the high level of empirical basis on how outcomes should be interpreted was also an important additional motivation aspect. This work has the goal of extracting interpretable, transparent models from selected opaque decision models via a new readability-enhanced multi-objective Genetic Programming (GP) approach. The proposed more interpretable decision models mimic the original OM, and yield similar classification outcomes for the same input data, while keeping model complexity low. Our proposition is grounded on the assumption that higher model complexity hinders interpretability. In light of that, we adapt text readability metrics into proxies to evaluate ML interpretability. Our results on benchmark data sets demonstrate that the readability-based metrics put forward are effective means for assessing interpretability when compared with the state-of-the-art approaches. Experimentally, we observed the practical ability of applying our approach, as we compared the results with our already-known competitors, considering that this study has used better reference of taking interpretability outcomes with a readable-enhanced evolutionary approach.
Divulgação - Defesa Nº 253
Aluno: Leonardo de Leon Dias
Título: “Sistema híbrido para previsão de velocidade do vento”
Orientador: João Fausto Lorenzato de Oliveira - (PPGEC)
Co-orientador: Emerson Alexandre de Oliveira Lima - (UPE)
Examinador Externo: Paulo Salgado Gomes de Mattos Neto - (UFPE)
Examinador Interno: Cleyton Mário de Oliveira Rodrigues - (PPGEC)
Data-hora: 28 de julho de 2022 às 13:00h.
Local: Formato Remoto (https://meet.google.com/frg-wdbr-bee)
Resumo:
A contribuição da energia eólica com a matriz energética mundial vem crescendo substancialmente e representa uma grande parcela na produção de energia limpa. Contudo, a capacidade de geração de energia eólica está diretamente relacionada à velocidade do vento, e o vento por sua vez é intermitente, apresenta variação constante, possui comportamentos variados e possui padrões não lineares. Dessa forma, a capacidade de prever a velocidade do vento é fundamental para viabilizar a instalação e operação de uma usina eólica. Diversos modelos e diferentes abordagens para previsão de séries temporais de velocidade do vento podem ser encontrados na literatura, dentre eles a modelagem de sistemas híbridos combinando diferentes modelos estatísticos mostram-se boas opções para o desempenho dessa tarefa. Esses sistemas visam superar as limitações de um único modelo através da agregação das qualidades oferecidas por modelos distintos. Neste sentido, a combinação de modelos lineares e não lineares para predição da série de resíduo é uma abordagem recorrente na literatura e que mostra-se bastante eficaz. Esse artigo propõe um sistema híbrido para previsão de séries de velocidade do vento com intervalos horários e mensais e utiliza diferentes modelos estatísticos para geração de uma função de combinação não linear entre modelos lineares e não lineares. A abordagem proposta orienta quais os modelos são mais adequados para proporcionar uma melhor performance na tarefa de predição. Foram feitas avaliações em diferentes cenários utilizando dados de três estações meteorológicas do nordeste do brasileiro e os resultados obtidos mostraram que o sistema híbrido proposto atingiu uma precisão superior a outros modelos encontrados na literatura.
Divulgação - Defesa Nº 252
Aluno: Marcos André Santos Galindo
Título: “Including Cultural Aspects in FSS - Cultural Weight-Based Fish School Search (cwFSS)”
Orientador: Fernando Buarque de Lima Neto - (PPGEC)
Co-orientador: Marcelo Gomes Pereira de Lacerda
Examinador Externo: Francisco Madeiro B. Júnior - (UNICAP)
Examinador Interno: Carmelo José Albanez B. Filho - (PPGEC)
Data-hora: 30 de junho de2022 às 16:30h.
Local: Formato Remoto (https://meet.google.com/ajd-cfpq-aah)
Resumo:
Many real-life engineering applications are optimization problems. To find the best combination of parameters to minimize costs and maximize efficiency, engineers tipically use project software such as CAD, CAE and CAM. In this context, intelligent optimization techniques may be used to automate and improve such task. None the less, these algorithms may search in unreliable areas and even suggest risky solutions, which for real-life applications may cause problems because of inaccuracies and unfeasibility (especially in the industrial environment). Thus cultural aspects should be combined with multimodal Swarm Intelligence algorithms to deal with that. The present work proposes the incorporation of Cultural Algorithms concepts to Weight-Based Fish School Search (wFSS), generating a new optimization algorithm, the Cultural Weight-Based Fish School Search (cwFSS). cwFSS is able to guide the optimization process based on norms, expert experience and theoretical knowledge about the problem without the need of implying constraints to the fitness function, while it keeps some desirable freedom to the search. cwFSS was also evaluated the use of historical knowledge to intelligently find the optimal time to stop the search process. The proposed method was tested in a function set of the CEC niching optimization competition, a thermal power plant efficiency optimization simulator, compared with the standard wFSS, and Niching Migratory Multi-swarm Optimiser (NMMSO) – champion of CEC’2015 niching optimization competition. As for results, cwFSS has outperformed NMSSO in time, fitness and variability, and the original wFSS about time, stability, safeness and variability of the multimodal solutions. Therefore, cwFSS is deemed an interesting support tool for engineering decisions problems.
O Programa
