| Grant number: | 18/15014-0 |
| Support Opportunities: | Scholarships in Brazil - Master |
| Start date: | March 01, 2019 |
| End date: | January 31, 2020 |
| Field of knowledge: | Physical Sciences and Mathematics - Computer Science - Computer Systems |
| Principal Investigator: | Lilian Berton |
| Grantee: | Bruno Klaus de Aquino Afonso |
| Host Institution: | Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil |
Abstract The recent growth of online content on the Web has allowed collecting more user-generated data with noisy and missing labels, e.g., data from social tags and voted labels from Amazon's Mechanical Turks. Most machine learning methods, which require accurate label sets, could not be trusted or have good performance. Graph classification has drawn increasing interests due to the rapid rising of applications involving complex network structured data with dependency relationships. There have been a number of studies on graph classification in recent years, especially in semi-supervised learning (SSL), since it works with a small amount of labeled data together with a big quantity of unlabeled data. These methods cannot be effective in the presence of noise (i.e. mislabeled samples) or outliers. For graph applications, due to the complexity of examining and labeling the structural networks, it is very difficult to obtain a completely noise-free dataset. So, effective designs for handling imbalanced and noisy graph data are highly desirable. In this work, we aim to propose new methods to handle label noise in graph-based SSL, encompassing label noise cleansing and label noise-tolerant methods. (AU) | |
| News published in Agência FAPESP Newsletter about the scholarship: | |
| More itemsLess items | |
| TITULO | |
| Articles published in other media outlets ( ): | |
| More itemsLess items | |
| VEICULO: TITULO (DATA) | |
| VEICULO: TITULO (DATA) | |