We consider the problem of distributed stochastic optimization, where each of several machines has access to samples from the same source distribution, and the goal is to jointly optimize the expected objective w.r.t. the source distribution, minimizing: (1) overall runtime; (2) communication costs; (3) number of samples used. We study this problem systematically, highlighting fundamental limitations, and differences versus distributed consensus problems where each machine has a different, independent, objective. We show how the best known guarantees are obtained by a mini-batched SGD approach, and contrast the runtime and sample costs of the approach with those of other distributed optimization algorithms, including distributed ADMM, Newton-like and quasi-Newton approaches.
Current 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996