A Comparison of the Primal-Simplex and Primal-Dual
Algorithms for Linear Programming

A statistical comparison of the primal-dual
and the more commonly used primal-simplex algorithm 
for solving linear programming problems has been made under
the assumption of starting with a full artificial 
basis.  Under these conditions the primal-dual method
shows a statistically significant superiority on 
randomly generated problems.  It has also been found, via
a regression analysis, that the relevant parameters 
in determining the difference in the number of iterations
between the algorithms is not only the number 
of constraints and the number of variables but
also the ratio of the latter to the former.

CACM November, 1965

Mueller, R. K.
Cooper, L.

CA651110 JB March 6, 1978  4:50 PM

1171	5	1171
1171	5	1171
1171	5	1171