vldb 2009

Finally, our paper on ranking over probabilistic datasets is accepted in VLDB. Since I like this paper very much, probly even more than any theory paper I have written so far, I just blog it here, for remembrance. To me, the ideas are neat and the algorithms are elegant, also very practical (I did a lot of experiments).  

This paper has been through a pretty tough time. The first draft was our report for the database course project. At that time, the paper was titled "clustering and ranking on prob databases" and devoted much space talking about clustering. We submitted it to vldb08 and got 2 weak accepts and one weak rej. Then the paper was rolloverred (a policy that some boarderline papers can be rolled to the next same level conf(sigmod/vldb) with the same set of reviewers) to sigmod09 and was reviewed by the same set of reviewers but got the same review results. 

Finally, we decide to completely remove the clustering part and focus only on the ranking part.

More results, explanations and experiment were added.  Especially, I added the ranking approximation part where I borrowed some ideas and results from signal processing (actually fourier transformation). This makes our new ranking function an very effective and efficient one.