Informações do Trabalho
Titulo
Deep Generative Models for Molecular Design
Subtítulo
Autor
LUÍS HENRIQUE SIMPLÍCIO RIBEIRO
Orientador
SAULO MORAES VILLELA
Resumo
Most of the effort spent today in material science and chemistry comes from the need to create molecules with specific properties, for instance, in medicine, the generation of molecules with desired properties plays a very important role in the discovery of new medicines. However, despite existing a finite number of molecules, this number is huge, making the exploration of this molecular space a hard problem of combinatorial optimization. Therefore, it is necessary to find a robust and efficient way to explore this molecular space. In this scenario, approaches like genetic algorithms had been used to try to handle the problem of generating molecules with specific properties computationally. More recently, Deep Generative Models, that are Machine Learning models capable of learning from data to generate new data that resemble the data they were trained on, are being used as an alternative to tackle the problem. With this work we present a detailed study of the complex problem of generating molecules using Deep Learning.
Ano:
2021
Palavras-Chave
Machine Learning, Deep Learning, Deep Generative Models, Molecular Generation.
Obter PDF
Obter arquivos extras
Obter Bibtex