Date: 06 July 2022
Time: 12:45 – 13:15. Hours
Room: RA1501 & online
Speaker: Hanyuan Hang (DAMUT-MOR)
Title: "Statistical Learning of Ensemble Methods”
Ensemble learning trains multiple learners to deal with machine learning problems. It has shown outstanding empirical performance and is widely applied in various fields of general science. A wealth of literature has paid attention to the theoretical and especially the experimental performances of ensemble algorithms, whereas there are a lot more to explore when meeting the new challenges of advanced and complex machine learning tasks. In my future research, I will investigate the convergence properties of ensemble learning under complex performance measures used in advanced tasks such as imbalanced classification and cost-sensitive learning. Moreover, I will show an interesting link between neural networks and ensemble learning, which enables us to provide new understandings of neural networks. Beside the complex supervised learning tasks, I will also investigate the effectiveness of ensemble learning in unsupervised and semi-supervised learning, which become increasingly important nowadays due to the hardness of getting high quality supervised data. Under these topics, new variants of parallel and sequential ensemble algorithms are to be proposed and analyzed under new theoretical frameworks of ensemble learning.