COPYRIGHT

I first want to thank my advisor, Dr. Daniel D. Lee, who offered me a fantastic opportunity to become a researcher in the field of machine learning. When I was starting my Ph.D. study in Spring 2003, I (literally) knew little about machine learning since my major had been optical engineering. It was very fortunate for me to have the chance to work with Dan. I am grateful for his knowledgable guidance and passionate support in the past five and half years. Without him, I would never be where I am. I would also like to thank my committee members. I am thankful to Dr. Jingdong Chen (Bell Labs) who was also one major collaborator of mine on our acoustic signal processing projects. He provided tremendous help with recognizing the frontier of acoustic signal processing research as well as conducting experiments in an anechoic chamber. My discussions with him were always pleasant and fruitful. I thank Dr. Lawrence Saul for helpful discussions and suggestions, as well as our collaboration on a couple of papers. Moreover, his lectures have been very beneficial to me, and I am glad that I took all the courses that he had taught at Upenn before he moved to UCSD. I also thank Dr. Saleem A. Kassam for chairing the committee and providing much helpful comments and suggestions. Besides Jingdong and Lawrence, I am thankful to my other collaborators during my Ph.D. study. I thank Dr. Youngmoo E. Kim (Drexel University) for allowing me to use the acoustic facility in his lab and beneficial discussions on advancing our acoustic signal processing projects. I also thank Dr. Bill Hanson and Dr. Erica Thaler (both from Hospital iii of the University of Pennsylvania) for involving me in the electronic nose project where I learned and implemented many classification algorithms. I would also like to thank Dr. Ben Taskar for our collaboration on learning sparse Markov networks, although this work is not included in this thesis. Meanwhile, I thank many of my wonderful colleagues and friends, Koby Crammer