ABSTRACT
Most traditional optimization methods used for manufacturing applications can be divided into two broad classes: direct search methods requiring only the objective function values and gradient search methods requiring gradient information either exactly or numerically. One common characteristic of most of these methods is that they all work on a point-by-point basis. An algorithm starts with an initial point (usually supplied by the user) and depending on the transition rule used in the algorithm, a new point is determined. Essentially, algorithms vary according to the transition rule used to update a point.
