Multi-objective optimization of a neural network-based nonlinear equalizer in unrepeated digital coherent optical communication systems

• Ano de defesa: 2024
• Autor(a) principal: Dantas, Lucas Cardinal [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Comunicações óticas; Inteligência artificial; Kerr, Efeito de
• Link de acesso: https://hdl.handle.net/11449/259327

In this work, we propose using neural networks to mitigate intersymbol interference introduced by the fiber in optical communication systems. This mitigation method is well known, nevertheless the majority of studies that we traced back in the literature process only the symbol of interest, and consequently, cannot compensate the intersymbol interference. The approach presented in this work includes also using adjacent symbols, as there is a correlation between these symbols distortion. But, this increases the computational complexity. For that reason, we also analyzed the computational complexity to train the neural network. We were not able to trace back in the literature works that performed this multi-objective analysis, taking into account the adjacent symbols for training the neural network and also the computational complexity.