[Coin-lpsolver] Querying the feasible bounds

[Lou Hafer]

Spencer,

	You can also try these:
	
  * Presolving in linear programming, Erling Anderson & Knud Anderson,
    Mathematical Programming v71, pp. 221-245, 1995.

  * Computational Techniques of the Simplex Method, Istvan Maros, Kluwer, 2003
    (Chap. 7)

Arc consistency can be applied to continuous or integer variables with no real
problem as long as you're willing to work with a single upper and lower bound.
The restriction to linear constraints makes the whole thing seem trivially
simple to the CSP (constraint satisfaction problem) people in AI.  Holes in the
range make the algorithms more complicated.  For LP/IP, a single [lb, ub] range
works just fine.  bonsaiG, an old IP code of mine, used arc consistency.
(http://www.cs.sfu.ca/~lou/BonsaiG).  A more sophisticated current approach is
SCIP (http://scip.zib.de/)

							Lou