Mengjie Zhang, Evolutionary Transfer Learning

Evolutionary Transfer Learning

Data mining, machine learning, and optimisation algorithms have achieved promises in many real-world tasks, such as classification, clustering and regression. These algorithms can often generalise well on data in the same domain, i.e. drawn from the same feature space and with the same distribution. However, in many real-world applications, the available data are often from different domains. For example, we may need to perform classification in one target domain, but only have sufficient training data in another (source) domain, which may be in a different feature space or follow a different data distribution. Transfer Learning aims to transfer knowledge acquired in one problem domain, i.e. the source domain, onto another domain, i.e. the target domain. Transfer learning has recently emerged as a new learning framework and hot topic in data mining and machine learning.

Evolutionary computation techniques have been successfully applied to many real-world problems, and started to be used to solve transfer learning tasks. Meanwhile, transfer learning has attracted increasing attention from many disciplines, and has been used in evolutionary computation to address complex and challenging issues.

The project aims to investigate transfer learning in evolutionary computation, covering ALL different evolutionary computation paradigms. The aim is to investigate in both the new theories and methods on how transfer learning can be achieved with different evolutionary computation paradigms, and how transfer learning can be adopted in evolutionary computation, and the applications of evolutionary computation and transfer learning in real-world problems.

Specifically, we would like to develop new algorithms for

  • Evolutionary supervised and unsupervised transfer learning.
  • Domain adaptation and domain generalization in evolutionary computation.
  • Transfer learning in evolutionary computation for classification, clustering, regression tasks.
  • Transfer learning in in evolutionary computation for scheduling and combinatorial optimisation tasks such as web service composition.
  • Transfer learning in in evolutionary computation for real-world applications such as image analysis.
A strong background in Java/C/C++ programming and a basic background in Artificial Intelligence and statistics are required. A good background in machine learning, and operations research is desired (COMP307, COMP361).

This project will be co-supervised by Dr Bing Xue. Please check,, and for publications and other information.