The multiplicative penalty function method for linear
programming is discussed for the cases (1) the optimum objective
function value is known in advance and (2) the value is not known
at the beginning.  Especially, the local convergence rate is
discussed for the case (1).