Compared with many others,
B&B procedure which uses the enumerative divide-and-conquer technique is a better method.
Besides, B&B has already been in common use now. However, it might require an enormous amount of computation sometimes when solving large-scale ILP problems. Although there are heuristics for enhancing the ability of B&B by guessing which branch could lead to a quick solution, there is no solid theory that will always yield consistent results Hence, in this section, we propose a revise method to IPL Our method to cutoriginal the space many subspaces using objective function before implementing It could narrow down feasible solution range. After labeling these subspaces based on the"distance" from objective function, we can apply upon subspaces one by one. The nearest subspace will be the first one to be searched The first optimum solution we get in the nearest sub is promised to be the optimum solution in all solution space because of using objective function. The procedure of our proposed i explained belo
e relaxed problem as usual without integer restriction to get the optimum point. Denote this point as A Step 2: dges, find the extreme points around point A and calculate the values of objective function ith these extreme points to determine which one is the nearest point. Then this extreme point is rep- resented as B 3: Add a new constraint which passes through B and parallels the objective function. Use original con straints and this new constraint to form two subspaces. These two subspaces are both smaller than ori solution space and label the area near A as the target subspace. Step 4: Searching for integer optimum solution using B&B procedure in this target subspace. If we can the integer optimum solution, then terminate and claim that this is the optimum point. If the integer optimum solution is not found in this subspace, then return to step 2 to continue looking for the next nearest subspace. Stop after the result is found This method can reduce the amount of constraints and make the computation more efficient. Because B&B usually uses generalized simplex algorithm to search for optimum solution in solution space, fewer constraints also mean less operations in each generalized simplex operation. Some constraints do not influence the out come of the computation; therefore, they could be removed from the constraint set. Our method will be of great usage while handling the ILP problems with many constraints. Besides, the proposed procedure could reduce the usage of computer memory. What if the ILP problems are not with many constraints but with many variables and few constraints? Then the dual theorem could be applied to transform the primal to its ual. The variables in the dual problem can be transferred to the constraints in the primal problem. In the same way, the constraints in the dual problem can be transferred to the variables in the primal problem. For the case many variables and less constraints, duality theorem should be used first to transform with constraints. Then our method can reduce the original problems to the case with fewer variables and many the number of constraints by shrinking the solution space