Botond SZABO (Leiden University) – “On the fundamental understanding of distributed computation”
April 1, 2:00 pm - 3:15 pm
The Statistical Seminar: Every Monday at 2:00 pm.
Time: 2:00 pm – 3:15 pm
Date: 1st of April 2019
Place: Room 3001.
Botond SZABO (Leiden University) – “On the fundamental understanding of distributed computation“
Abstract: In recent years, the amount of available information has become so vast in certain fields of applications that it is infeasible or undesirable to carry out the computations on a single server. This has motivated the design and study of distributed statistical or learning methods. In distributed methods, the data is split amongst different administrative units and computations are carried out locally, in parallel to each other. The outcome of the local computations are then aggregated into a final result on a central machine.
We consider the limitations and guarantees of distributed methods under communication constraints (i.e. only limited, fixed amount of bits are allowed to be transmitted between the local and the central machines) in context of the random design regression model. We derive minimax lower bounds (which depending on the communication budget can be substantially higher than the standard non-distributed minimax rates), matching upper bounds and provide adaptive estimators reaching these limits. We also consider the case where the number of transmitted bits is taken to be data-driven and investigate whether one can achieve the minimax non-distributed estimation rate and at the same time transmit in some sense the optimal amount of information between the machines.
This is a joint work with Harry van Zanten.
Cristina BUTUCEA, Alexandre TSYBAKOV, Julie JOSSE, Eric MOULINES, Mathieu ROSENBAUM