The OPTLP Procedure

Pricing Strategies for the Primal and Dual Simplex Algorithms

Several pricing strategies for the primal and dual simplex algorithms are available. Pricing strategies determine which variable enters the basis at each simplex pivot. They can be controlled by specifying the PRICETYPE= option.

The primal simplex algorithm has the following five pricing strategies:

PARTIAL

uses Dantzig’s most violated reduced cost rule (Dantzig, 1963). It scans a queue of decision variables and selects the variable with the most violated reduced cost as the entering variable. You can optionally specify the QUEUESIZE= option to control the length of this queue.

FULL

uses Dantzig’s most violated reduced cost rule. It compares the reduced costs of all decision variables and selects the variable with the most violated reduced cost as the entering variable.

DEVEX

implements the Devex pricing strategy developed by Harris (1973).

STEEPESTEDGE

uses the steepest-edge pricing strategy developed by Forrest and Goldfarb (1992).

HYBRID

uses a hybrid of the Devex and steepest-edge pricing strategies.

The dual simplex algorithm has only three pricing strategies available: FULL, DEVEX, and STEEPESTEDGE.