We propose two different crash-starting technique for interior point methods applicable to stochastic programming problems.
In the first case a series of simplified problem with fewer scenarios (obtained from scenario reduction) are solved to successively obtain an estimate of an advanced point on the central path of the full problem. In the second case only the first stage components of such an advanced central point are obtained by performing half an iteration of a decomposition scheme on the barrier problem correponding to the deterministic equivalent.
We analyse the conditions under which such a scheme is successful, argue that it leads to improved complexity and give numerical results obtained by the IPM solver OOPS.
Current 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996