Some Machine Learning Methods from Sequential Input

First, I would like to thank my advisor Professor Xiaojin (Jerry) Zhu. His enthusiasm, relentless pursuit of excellence, and ability to examine problems from every single angle were a great inspiration for me. It was Jerry who gave me an opportunity to start research in the first semester of my graduate program and warmed me up with a nonparametric Bayesian model. He happily let me read his books, and sat down with me to explain the math. He also taught me how to debug mathematical codes. He cared so much about my career and experience that he encouraged me to apply for summer internships. The industrial job experience and the networks I made through these internships are now an invaluable asset to me. Whenever I needed to give a talk at an important place, his feedbacks on my presentations were truly a million dollar lesson. He also invited me to numerous meetings with other researchers, which became a great opportunity to collaborate. Thanks to him, I met Professor Tim Rogers, Professor Robert Nowak, Kevin Jamieson, and Professor Ming Yuan who all became my coauthor. He also encouraged me to start a reading group on non-iid learning, which became a firm base for my research. Organizing and leading this group were a priceless experience.More than half of my thesis could not have happened without Professor Timothy Rogers. He introduced me to the world of cognitive psychology. He happily shared his data and analysis for our research in a neat format, and provided insight on human cognition. Sections he wrote for publications were also a great lesson on academic writing.