0% found this document useful (0 votes)
3 views6 pages

En Sample ExamQuestions2

The document presents a series of problems related to branch-and-bound algorithms for minimization problems with binary and integer variables. It includes specific questions regarding the elimination of subproblems based on bounds, the determination of best bounds after processing nodes, and the calculation of maximum absolute errors. Additionally, it discusses the application of the branch-and-cut method and the identification of valid inequalities to improve objective function values in the context of solving linear programming relaxations.

Uploaded by

samadani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
3 views6 pages

En Sample ExamQuestions2

The document presents a series of problems related to branch-and-bound algorithms for minimization problems with binary and integer variables. It includes specific questions regarding the elimination of subproblems based on bounds, the determination of best bounds after processing nodes, and the calculation of maximum absolute errors. Additionally, it discusses the application of the branch-and-cut method and the identification of valid inequalities to improve objective function values in the context of solving linear programming relaxations.

Uploaded by

samadani
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

PROBLEM 1: (20 points)

The following figure represents branch-and-bound tree obtained during the solution of a
minimization problem where the variables are all binary. Recall that the subproblems
obtained from a problem given by branching to a fractional variable are called child
nodes and that the problem which gave rise to them is referred to as parent node.
The numbers in the vertices identify the subproblem to which that node corresponds and
indicate the order in which the subproblems were solved. Vertex 0 indicates the root
node. The values appearing next to the vertices indicate the optimal value of the objective
function of the linear relaxation of the corresponding subproblem.

The linear relaxation associated with the subproblem 4 has just been solved and the
optimal solution obtained is the first integer solution found since the start of the solution.
Nodes a, b, c, and d have not yet been processed.
Question 1: If the value of the objective function corresponding to the integer solution
found at node 4 is 123, indicate which subproblems could be eliminated (based on the
bound associated with their parent nodes).
Question 2: Once node 4 has been processed, determine what is the best bound on the
value of the objective function of the optimal solution of the integer problem that is
available at this time.
Question 3: At this point in the solution, calculate what would be the maximum absolute
error if one decided to stop after getting the integer solution at node 4.

Answer 1: The objective function is a minimization. Therefore, an integer solution with


an objective function value of 123 provides an upper bound (we can possibly improve
this value (get smaller objective function values), but this is the best we have achieved so
far). Considering the fact that node 3 has an LP relaxation of 125, nodes b and c can be
pruned by bound. Nodes a and d still have the promise to improve the objective function
value, therefore they need to be further explored.

1
Answer 2: Since node d is not still explored, the best bound on the right branch of the
root node is 118. Similarly, node a is not explored yet, therefore the best bound on the
left branch is 120. Therefore, the best bound is at the moment 118.

Answer 3: When the solution process is terminated after node 4 is explored and an
integer-feasible solution with an objective function value of 123 is obtained, then the
maximum absolute error is best integer – best bound = 123-118 = 5 and the maximum
118 = 4.2%.
relative error is 5/128

PROBLEM 2 : (20 points)


We want to solve the problem
minimize f (x1, x2) = –x1 – 5x2
subject to 2x1 + 2x2 ≤ 7
–x1 + 2x2 ≤ 2
x1, x2 ≥ 0 and integer
whose continuous relaxation is x1 = 5/3, x2 = 11/6 as an optimal solution. The graphic
solution of this problem is below.

Question 1: Solve this problem by the branch-and-bound (BB) algorithm using the
breadth-
depth-first search strategy and branching on the most fractional variable.

Solve the subproblems of your BB tree graphically using the figure provided on the next
page (considering that the horizontal axis represents the values of x1 and the vertical axis
those of x2). Report the solutions of the subproblems on a BB tree.

2
Question 2 Provide a valid inequality for this problem.

3
Answer 1:

P0

P1 P2

P3 P4 P5 P6

P7 P8

Answer 2: The cutting plane (the valid inequality) should remove the optimal objective
function value of the linear programming relaxation. Below are 3 such valid inequality
examples:

𝑥𝑥2 ≤ 1 (from the graph)

𝑥𝑥1 + 𝑥𝑥2 ≤ 3 (2nd constraint is 2𝑥𝑥1 + 2𝑥𝑥2 ≤ 7 ⟺ 𝑥𝑥1 + 𝑥𝑥2 ≤ 3.5 ⟺ 𝑥𝑥1 + 𝑥𝑥2 ≤ 3)

𝑥𝑥1 + 2𝑥𝑥2 ≤ 4 (from the graph)

4
PROBLEM 3 : (20 points)
The following figure represents the BB tree summarizing the solution of a minimization
problem where the variables are binary. "Branch-and-cut" method is used as the solution
algorithm. The solutions obtained by solving the continuous relaxations of each problem
as well as the value of the corresponding objective function are indicated to the right of
the node representing this problem.

Briefly describe the flow of the algorithm schematized by this tree by explaining how and
why nodes were "branched", the valid inequalities, when integer solutions were found,
and which solution was declared optimal (We assume that all the added inequalities are
valid for the starting problem).

Answer: The LP relaxation in the root node is solved and a fractional solution is obtained
with an objective function value of 50. In the second stage, a valid inequality is identified
that removes the previous LP relaxation solution. Appending x2 + 2x3 ≥ 2 to the model
strengthened the formulation and the LP relaxation objective function value is improved
to 57. The same process is repeated as the third stage and a new inequality is added to
improve the objective function value to 58. At this stage, we branch on a fractional
variable x1. The left branch enforces x1=1 and the right branch enforces x1 = 0. Solving
the left branch, we obtain an integer feasible solution with an objective function value of
63, which is the first integer feasible solution encountered in the search tree and it

5
provided an upper bound (since the model is a minimization). This branch is pruned by
integrality. The LP relaxation on the right branch is then solved and a fractional solution
with an objective function value of 61 is obtained. A new valid inequality is identified as
the next step at node 5 to remove the previous optimal solution to the LP relaxation and
upper bound at hand, which is 63 and
bound is improved to 65. This exceeded the best lower
therefore we prune this branch by bound. Since all unexplored nodes are explored, the
algorithm terminates with an optimal objective function value of 63 and optimal solution
is (1, 0, 1).

You might also like