Adapting self-organizing maps to the MapReduce programming paradigm 123
The MapReduce programming paradigm first proposed by J. Dean and S.
Ghemawat (see [6]) provides a framework for designing and implementing
distributed algorithms. It was primarily intended for use at Google Inc. to
handle their huge amount of data (according to [6] more than 20 petabytes
a day). Within a MapReduce context, the framework takes care of failure-
correction, data distribution and load management thus taking off a great share
of issues from the programmers shoulders.
In MapReduce generally data is present as key-value pairs (k, v) ∈ K
i
×V
i
.
As the name implies the M/R