An asymptotic analysis of nonparametric divide-and-conquer methods
In the era of “big data” computer scientists and statisticians are facing new challenges to handle the large amount of available information. Datasets have become so large that it is unfeasible, or computationally undesirable, to carry out the analysis on a single machine. This gave rise to divide-and-conquer (also known as distributed) algorithms where the data is distributed over several “local” machines and the computations are done on these machines parallel to each other. Then the outcome of the local computations are somehow aggregated to a global result in a central machine.
In this talk I will investigate various distributed methods in a unified, benchmark, nonparametric statistical model and derive asymptotic guarantees and limitations for these procedures.
This is a joint work with Harry van Zanten.