Associate ProfessorChang Xu

Nowadays, data is increasingly cheap and ubiquitous. The corpuses are too voluminous to fit on a single machine. Meanwhile, the variety of data stands out on its own, as the data often include various data types, such as video, music and text. They are heterogeneity in modality, sources and semantics. On the other hand, they are typically messy, incomplete, and unstructured. Given these challenges, the ability to handle data variety and use it to our advantage is becoming ever more important.

Dr Chang Xu's passion is to devise ways to understand how data variety works and how it can be interrogated and exploited for the broadest possible purposes.

"We humans can intelligently integrate information from their natural senses of vision, hearing, touch and so on to make the decision. This innate ability enables us to confidently interpret our current environment or situation no matter how complex it is.

"A machine can also easily collect diverse information through different sensors or extractors, e.g., colour descriptor, local binary patterns, local shape descriptor and spatial temporal context captured by multiple cameras for human activity understanding in camera network, or the word content of webpage, and the website address and name to distinguish this particular web page before your eyes now.

"How best to describe these heterogeneous data, explore their underlying properties and inner relationships, and leverage them for distinct decision needs excite me about the research.

"I identify significant challenges, provide fundamental theoretical supports, and formulate innovative algorithmic solutions to understand and manipulate data variety. One of my ultimate research aims is allowing machines to intelligently handle multiple information sources like humans.

"I warmly welcome students and peers to collaborate with me in the related research fields to make contributions to the realization of Artificial Intelligence."