Rafael Cruz, da École de Technologie Supérieure (ETS), ministrará a palestra no dia 23/08, às 15h, no auditório CIn/Samsung
O professor Rafael Cruz, da École de Technologie Supérieure (ETS), ministrará na próxima sexta-feira (23 de agosto) um seminário de pesquisa sobre a detecção de notícias falsas no Centro de Informática (CIn) da UFPE. Intitulada “Fake news detection: a dynamic multi-view and ensemble selection approach”, a palestra será apresentada em português. O professor é Ph.D da ETS Canadá irá discorrer sobre o desafio de identificar fake-news e como a mudança de abordagem de modelos e conteúdos para além de um problema textual pode facilitar essa compreensão. O seminário será realizado no Auditório do CIn/Samsung, no primeiro andar do Bloco B, a partir das 15h.
Sobre a palestra – This presentation tackles the challenge of textual fake news detection with a dynamic multi-view and ensemble selection approach. The discussion begins with an overview of fake news detection as a text classification problem, emphasizing the complementarity between feature representations and classification models. Our analysis shows that classifiers trained on different feature representations often make dissimilar errors, suggesting that using multiple feature spaces can significantly improve detection accuracy. Then, we propose a dynamic multi-view and ensemble selection (DMES) model that uses instance hardness to guide the selection of feature representations, adapting to the complexity of each newly generated content. By choosing the most appropriate representation for each input and integrating multiple classifiers, our approach enhances detection accuracy, obtaining a significant performance compared to the state-of-the-art method. We also analyze the view selection and ensemble selection steps to understand the limitations of the proposed DMES approach and possibilities for future improvements. The presentation will also include an overview of my research at the LIVIA laboratory, covering projects like 3D facial animation from audio, cross-modal learning, and dynamic ensemble selection.
Sobre o palestrante – Rafael M. O. Cruz received an M.Sc. in 2011 from the Federal University of Pernambuco (UFPE) and a Ph.D. in Engineering in 2016 from the École de Technologie Supérieure (ÉTS), Montréal, Quebec, where he is currently an Associate Professor. He is also a member of the Laboratoire d’imagerie, vision et intelligence artificielle (LIVIA) at ÉTS-Montréal. His research interests include ensemble learning, dynamic ensemble models, meta-learning, and social media mining.
Comentários desativados