Rong Ge

I'm a 5th year graduate student at the Computer Science Department of Princeton University. My research area is Theoretical Computer Science. My advisor is Sanjeev Arora.

Recently I've been working on learning problems. I am generally interested in the meta problem: machine learning people have been applying heuristic algorithms such as local search or EM type algorithm to find solutions to many hard problems. Are these problems really easy or hard? If they are easy can we use algorithms with provable guarantees; if they are hard is there any reasonable model where they become easy or even the known heuristic algorithms can be proved to work?