The award committee considered Dr. Ma’s dissertation entitled “From Concepts to Events: A Progressive Process for Multimedia Content Analysis” worthy of the recognition as the proposed framework based on mathematical theories has great potential for developing real-world applications as well as addressing myriad technical challenges.
The fundamental innovations presented in Dr. Ma’s thesis consist of
- feature selection through subspace sparsity which leads to greatly improved accuracy with compact representation,
- semi-supervised learning with joint feature selection allowing exploitation of massive unlabeled data with only few labeled data,
- multimedia event detection by learning an intermediate representation,
- knowledge adaptation for multimedia event detection when only very few examples are available.
Despite the variety of problems addressed, these innovations are based on a unified machine learning framework, which is applicable to diverse application domains. The proposed solutions have been proven to be effective and general through a large set of experiments over a variety of challenging data sets, including personal photos, web images, consumer videos, You Tube style internet video corpora, health care surveillance data, and 3D human motion data.
Bio of Awardee
Dr. Zhigang Ma received the B.Sc. and M.Sc. degrees from Zhejiang University, China, in 2004 and 2006, respectively, and the Ph.D. degree from University of Trento, Italy, in 2013. The title of his thesis is “From Concepts to Events: A Progressive Process for Multimedia Content Analysis”. He is currently a Postdoctoral Research Fellow with the School of Computer Science, Carnegie Mellon University, Pittsburgh, USA. His research interest is mainly in multimedia analysis using machine learning techniques. He received the best PhD thesis award from Gruppo Italiano Ricercatori in Pattern Recognition, Italy, in 2014. He was a PC member for ACM MM 2014 and a TPC member for ICME 2014.