Um modelo Autoencoder para lidar com dados ausentes e com ruído

• Grant number: 23/13688-2
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• Effective date (Start): March 15, 2024
• Effective date (End): September 14, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Ana Carolina Lorena
• Grantee: Arthur Dantas Mangussi
• Supervisor: Pedro Henriques Abreu
• Host Institution: Divisão de Ciência da Computação (IEC). Instituto Tecnológico de Aeronáutica (ITA). Ministério da Defesa (Brasil). São José dos Campos , SP, Brazil
• Research place: Universidade de Coimbra (UC), Portugal
• Associated to the scholarship: 22/10553-6 - An unified approach for dealing with missing and noise data, BP.MS

Abstract

Various issues can deteriorate data quality in Machine Learning. Among them, one may cite the quality of the input features, which can be impaired by the presence of noise and missing data. Noise can be inputted into data in several steps during collection, storage, and transmission. In supervised problems, they can be present both in the labels and in the predictive input features. Missing values are also a concern, as other information about the data item can be essential and must be regarded. This project will investigate these two problems and their interplay, including strategies to deal with them by taking possible corrective approaches. The objective is to experiment with different Autoencoder algorithms proposed by the Portuguese research group for imputing missing data, which are currently state-of-the-art in this area, and extend these algorithms to also identify and potentially correct noisy data.