mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <ted.dunn...@gmail.com>
Subject Re: page rank algorithm?
Date Thu, 01 Jul 2010 18:17:05 GMT
Jimmy Lin's presentation (first link on this page:
http://www.umiacs.umd.edu/~jimmylin/) had to do with data structure
improvements for link distance computations.  After his talk, there was an
interesting discussion with Arun Murthy of the map-reduce team at Yahoo.

Arun's contention was that it would be quite possible to have the combiner
do essentially the same work as Jimmy's data structure.  Jimmy was quite
dubious that the speedups would come for free and had portability concerns
about the speedup related to control issues (i.e. he controls his own clever
structures while the combiner might lose interest in helping him on later
versions of Hadoop).  Arun was unconvinced by Jimmy's protestations and
Jimmy was dubious that a general purpose mechanism could do as well.  They
agreed to disagree and Jimmy suggested that Arun implement a graph search
using a combiner so they could test.  Given Arun's massive workload, that
isn't likely.

On Thu, Jul 1, 2010 at 11:02 AM, Jake Mannix <jake.mannix@gmail.com> wrote:

> What were Jimmy's improvements, Ken?  Improvements to the algorithm, or the
> implementation, or what?
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message