时间：7月4日上午10:00

地点：行政楼C505

主讲人：Wei Liu

Abstract: 

Designing effective and efficient indexing schemes and search algorithms has recently attracted considerable attention due to the explosive growth of data, such as Web documents, images, and consumer videos on the Internet. Since exact nearest neighbor search is infeasible for large-scale multimedia applications which require exhaustive data scanning and huge memory overhead, hashing based approximate nearest neighbor (ANN) search has become popular owing to its practical efficiencies in both storage and search. Randomized hashing methods, e.g., locality sensitive hashing (LSH) and Min-Hash, explore data-independent hash functions leveraging random projections and permutations. In spite of theoretic guarantees in terms of approximating nearest neighbors for certain distance or similarity metrics, the resulting indexing is often less accurate and efficient in many practical applications. This is because of the fact that semantically similar samples cannot be easily mapped to adjacent binary codes by randomized hash functions. Recently, tremendous efforts have been paid to design more efficient and semantics-aware hashing techniques though incorporating various machine learning tools and algorithms to yield compact binary codes, which are named learning to hash in literature. In this tutorial, we provide a comprehensive survey of the recent developments of learning to hash in both methodologies and applications, ranging from unsupervised to supervised approaches. In particular, we compare the motivations, objectives, and solutions of popular learning-based hashing methods and also discuss the pros and cons. Finally, we will discuss the future directions and trends of hashing being applied to large-scale multimedia search. 

Bio: Dr. Wei Liu received the M.Phil. and Ph.D. degrees in electrical engineering from Columbia University, New York, NY, USA in 2012. Currently, he is a research staff member of IBM T. J. Watson Research Center, Yorktown Heights, NY, USA, and holds an adjunct faculty position at Rensselaer Polytechnic Institute, Troy, NY, USA. He has been the Josef Raviv Memorial Postdoctoral Fellow at IBM T. J. Watson Research Center for one year since 2012. His research interests include machine learning, data mining, computer vision, pattern recognition, multimedia, and information retrieval. Dr. Liu is the recipient of the 2011-2012 Facebook Fellowship and the 2013 Jury Award for best thesis of Department of Electrical Engineering, Columbia University. Dr. Liu has published over 60 papers in peer-reviewed journals and conferences such as Proceedings of IEEE, IEEE Transactions on Image Processing, ICML, KDD, CVPR, ICCV, ECCV, MICCAI, ACM Multimedia, IJCAI, AAAI, SIGIR, SIGCHI, etc. His recent paper wins the best paper travel award for ISBI 2014.