Fri, 24 Apr 2009 12:23:17 +0200Support LEQ and GEQ supply constraints in dimacs-solver (#234, #219)
Peter Kovacs <kpeter@inf.elte.hu> [Fri, 24 Apr 2009 12:23:17 +0200] rev 614
Support LEQ and GEQ supply constraints in dimacs-solver (#234, #219)

Fri, 24 Apr 2009 12:22:06 +0200Bug fix in NetworkSimplex (#234)
Peter Kovacs <kpeter@inf.elte.hu> [Fri, 24 Apr 2009 12:22:06 +0200] rev 613
Bug fix in NetworkSimplex (#234)

Thu, 23 Apr 2009 10:44:35 +0100Fix usage of sqrt() (#268)
Alpar Juttner <alpar@cs.elte.hu> [Thu, 23 Apr 2009 10:44:35 +0100] rev 612
Fix usage of sqrt() (#268)

Tue, 21 Apr 2009 15:18:54 +0100Merge and fix
Alpar Juttner <alpar@cs.elte.hu> [Tue, 21 Apr 2009 15:18:54 +0100] rev 611
Merge and fix

Fri, 17 Apr 2009 18:14:35 +0200Slightly modify the interface of Circulation and Preflow (#266)
Peter Kovacs <kpeter@inf.elte.hu> [Fri, 17 Apr 2009 18:14:35 +0200] rev 610
Slightly modify the interface of Circulation and Preflow (#266)
in order to synchronize them to the interface of NetworkSimplex.

Circulation:
- The "delta" notation is replaced by "supply".
- lowerCapMap(), upperCapMap() are renamed to lowerMap() and upperMap().
- Value is renamed to Flow.

Preflow:
- Value is renamed to Flow.

Fri, 17 Apr 2009 18:04:36 +0200Support >= and <= constraints in NetworkSimplex (#219, #234)
Peter Kovacs <kpeter@inf.elte.hu> [Fri, 17 Apr 2009 18:04:36 +0200] rev 609
Support >= and <= constraints in NetworkSimplex (#219, #234)

By default the same inequality constraints are supported as by
Circulation (the GEQ form), but the LEQ form can also be selected
using the problemType() function.

The documentation of the min. cost flow module is reworked and
extended with important notes and explanations about the different
variants of the problem and about the dual solution and optimality
conditions.

Fri, 03 Apr 2009 18:59:15 +0200Support real types + numerical stability fix in NS (#254)
Peter Kovacs <kpeter@inf.elte.hu> [Fri, 03 Apr 2009 18:59:15 +0200] rev 608
Support real types + numerical stability fix in NS (#254)

- Real types are supported by appropriate inicialization.
- A feature of the XTI spanning tree structure is removed to ensure
numerical stability (could cause problems using integer types).
The node potentials are updated always on the lower subtree,
in order to prevent overflow problems.
The former method isn't notably faster during to our tests.