Speaker: Professor He Xiaofei (Zhejiang University)
Time: 9:30 am, December 13, 2010
Venue: Conference Room of Optical Image Analysis and Learning Center (OPTIMAL), 3rd Floor, Building 3

Summary:
Manifold learning is an emerging branch in the field of machine learning, which aims to study the data through manifold geometry and topological structure, and solve the problems of traditional machine learning, such as feature extraction, clustering, classification and so on. Manifold learning is a cutting-edge discipline using multi-disciplinary knowledge including differential geometry, graph theory, algebraic topology and the probability statistics to conduct data analysis based on manifold hypothesis. The report includes the following sections:
1. The rapid development of manifold learning in the decade, and some basic knowledge, such as the concept of manifold, manifold Laplace operator, and so on.
2. Classical manifold learning algorithms, such as Isoamp, Locally Linear Embedding, LaplacianEigenmap etc.; the semi-supervised learning and active learning algorithms based on manifold theory.
3. Extensive application of manifold learning in various disciplines, such as information processing.
4. The future of manifold learning, as well as the opportunities and challenges we face.

Speaker Profile:
He Xiaofei is a professor of Computer Science College at Zhejiang University, doctoral supervisor, and IEEE Senior Member. He received Bachelor’s Degree in Computer Science College at Zhejiang University, and had won the Grand Prize of International Mathematical Contest in Modeling (INFORMS). He got doctorate at University of Chicago. His doctoral dissertation proposed the first international linear manifold learning algorithms – Locality Preserving Projections (LPP), which set off a research boom of linear dimension reduction algorithm based on spectral graph theory. From 2001 to 2004, he and his colleagues in Microsoft Research Asia had been conducting information retrieval research. They proposed the link analysis method based on web block structure, which caused a great response in academia and industry. A number of international professional media reported this research. In 2005 manifold learning theory is introduced in face recognition and face method of Laplace was proposed. In 2008, he accepted an interview with Thomson Reuters, and his paper was cited as "the 14th most frequently cited papers in the world championship during the past decade” in face recognition. According to the statistics of Documentation and Information Center at Chinese Academy of Sciences, this paper was listed the second in Chinese scholars’ cited papers in engineering technology research papers from 2005 to 2009. In October 2005, he joined the United States Institute of Yahoo, serving as a researcher, being responsible for ad-search related research. In 2007, he was introduced in State Key Laboratory of CAD & CG at Zhejiang University as a talent, in which he established the research team in machine learning and information retrieval. In recent years, he published more than 60 academic papers and filed 7 U.S. patents, 3 of which have been authorized. His papers were cited a total of more than 2,800 times by others. The recommended work in the music received Best Paper Award nomination in International Conference on Multimedia 2010.