Niketan Pansare, Phd Student @ Rice university

No awards or grants added yet.

At a very high level, I work in the field of Statistical Databases and Data Mining; and hence most of my research projects involve building systems for performing statistical analysis on large scale data. To address my research problems, I generally resort to Bayesian methodology and use scalable systems (like Hadoop, Hyracks and Datapath).

A key problem that I am very excited about is providing online approximate answers on Map-Reduce systems. Though the problem of Approximate Query Processing has been studied extensively in the field of databases using sampling theory (See OLA97, Ripple Join and DBO), these traditional sampling techniques cannot be applied directly to Map-Reduce systems due to certain challenges (like inspection paradox). These are addressed in my VLDB paper, which also discusses a bayesian model to deal with these challenges.