I'm a PhD student in theoretical computer science at University of Michigan. My advisor is Seth Pettie.
Previously I am working on conflict resolving algorithms for contention resolution. Now I am working on cardinality estimation using stream algorithms. Both are still under investigation.
My ambition is to answer abstract but important problems in algorithm design. To device some algorithms is only of medium interest to me. I always desire to know what the optimal algorithm is. However, for some problems, there seem not to be a trivial way to define the exact "goodness" of an algorithm. Anyway, there are lots of books to be read, lots of things to be thought of and lots of work to be done.
What is cardinality estimation? The key point is to use a small amount of memory to estimate the number of unique elements in a long long stream. I focus on the class of algorithms using commutative and idempotent memory access schemes. On the one hand, I am studying the problem that, given a family of memory access functions, what the best probability assignments are and what the best estimator is. On the other hand, which is more interesting, for the class of natural independent-counter-group schemes, what is the best combination? And are there any other access schemes? If so, how does it compare to the the natural ones?
A small platform game: Moment Master.