Seminar on Manifold, clusters, and landmarks: exploiting intrinsic structures of data for unsupervised domain adaptation by Professor Fei Sha

  • Posted on: 16 May 2014
  • By: hadmin

Title: Manifold, clusters, and landmarks: exploiting intrinsic structures of data for unsupervised domain adaptation

Speaker: Professor Fei Sha
                 Assistant Professor
                 Department of Computer Science
                 University of Southern California

Date: 6 June2012
Time: 4:30pm – 5:30pm 
Venue: Room 215William M. W. Mong Engineering Building, CUHK

Statistical learning algorithms often assume the training data and the test data are drawn from the same (unknown) distribution. While this assumption facilitates rigorous theoretical analysis and empirical comparison, its validity is often challenged in real-world problems, where the testing environment of a learning agent frequently deviates from the well-controlled setting in which the agent is trained.Domain adaptation techniques aim to correct the mismatch between these two conditions. The goal is to adapt the classifiers trained on the source domain to work well in the test (target) domain.In this talk, I will present three vignettes of our recent work in developing new learning methodologies for domain adaptation. Our work examines and exploits intrinsic structures of the data: manifold, (discriminative) clusters and landmarks (instances connecting the source and target domains). I will describe those modeling assumptions and the corresponding learning algorithms for adaptation. I will use visual object recognition and sentiment analysis as application examples to validate our approaches. Our empirical results show that the proposed algorithms attain state-of-the-art performances on standard benchmark tasks.This talk is based on joint works with my students Boqing Gong and Yuan Shi, as well as our collaborator Prof. Kristen Grauman (U. of Texas, Austin).

About the speaker:
Fei Sha is an assistant professor at the University of Southern California, Dept. of Computer Science. His primary research interests are machine learning and application to speech and language processing, computer vision, robotics and others. After obtaining his PhD under the supervision of Prof. Lawrence K. Saul from U. of Pennsylvania, he worked as a postdoc with Profs. Michael I. Jordan and Stuart Russell at U. of California (Berkeley). 
He wrote his doctoral thesis on large margin based parameter estimation techniques for hidden Markov models. He has also worked extensively in dimensionality reduction. He has won outstanding student paper awards at NIPS and ICML. He is a member of DARPA 2010 Computer Science Study Panel, and won an Army Research Office Young Investigator Award (2012).

Inquiries: Alice Lau

Wednesday, June 6, 2012 - 16:30 to 17:30