Optimization from fundamentals: Week 7 (2022)

Question 1. Consider the following linear programming problem

maximize 3x1 + 2x3

x1 , x2 , x3
subject to x1 + x2 + x3 ≤ 6,
2x1 − x2 + x3 ≤ 3,
3x1 + x2 − x3 ≤ 3,
x1 , x2 , x3 ≥ 0

(0, 1.5, 4.5) is the optimal solution of this problem. The dual is given by:

minimize 6y1 + 3y2 + 3y3

y1 , y2 , y3
subject to y1 + 2y2 + 3y3 ≥ 3,
y1 − y2 + y3 ≥ 0,
y1 + y2 − y3 ≥ 2,
y1 , y2 , y3 ≥ 0

Let (y1∗ , y2∗ , y3∗ ) be the optimal solution of the dual problem. Then y1∗ = , y2∗ = and y3∗ = (fill
in the blanks with integers)

Solution: (y1∗ , y2∗ , y3∗ ) = (1, 1, 0)

Explanation: As 3x1 + x2 − x3 ≤ 3 does not hold with equality, y3 is 0 by complementary slackness.
Similarly, the second and third constraint of dual problem should hold with equality. Those are
y1 − y2 = 0 and y1 + y2 = 2. Solving y1 = 1, y2 = 1.

Question 2. Consider the shortest path problem with m nodes and n edges/arcs. The LP formulation
(discussed in lecture) can be written as

minimize cT x
x
subject to Ax = b,
0≤x≤1

where c is the cost of each edge/arc and A ∈ Rm×n . Which of the following is true about A?
(a) Every row has exactly one +1 and one −1, all others are zero.
(b) Every column has exactly one +1 and one −1, all others are zero.

Solution: (b) Every column has exactly one +1 and one −1, all others are zero.
Explanation: Each variable xij is for the edge/arc from node i to node j. So for corresponding
column of A, it will appear with −1 in i-th row (outflow) and with +1 in j-th row (inflow). All
other rows in that column will be zero.
Question 3. State TRUE or FALSE: For a primal and dual pair of a linear programming problems, if
some point satisfies complementary slackness condition then it is the optimal solution.
(a) TRUE (b) FALSE

Solution: (b) FALSE

Explanation: Feasibility is also required.
   
2 1 1 6
Question 4. State TRUE or FALSE: Consider A = 5 2 1 and b = 8. There exists a y such
3 2 3 0
that y T A = 0 (Zero vector) and y T b 6= 0.
(a) TRUE (b) FALSE
 Solution: (a) TRUE
Explanation: with y = (−4, 1, 1)T we have y T A = (0, 0, 0) and y T b = −16 6= 0

Question 5. Suppose you are given a LP to solve. You are able to formulate the dual of the problem.
But you are unable to solve both the problems. Then your friend comes along and informs you that x∗
is the optimal solution for primal problem. How do you check if your friend is correct?
(a) Check if x∗ is primal feasible
(b) Check if x∗ is primal feasible and find a dual feasible point y such that complementary slackness
holds
(c) No need to check, your friend is correct.
(d) Check if x∗ is primal feasible and find any point y such that complementary slackness holds

Solution: (b) Check if x∗ is primal feasible and find a dual feasible point y such that complementary
slackness holds.
Explanation: Refer to the complementary slackness theorem.

Question 6. State TRUE or FALSE: Assume that for a LP both primal and and dual is feasible. Then
both the problems have a finite optimal value.
(a) TRUE (b) FALSE

Solution: (a) TRUE

Explanation: If primal is unbounded the dual is infeasible. As the dual is feasible it means the
primal is bounded, hence have finite optimal value. Same argument hold for dual.

Question 7. State TRUE or FALSE: For an LP if primal is unbounded then dual is also unbounded
(a) TRUE (b) FALSE

Solution: (b) FALSE

Explanation: If primal is unbounded the dual is infeasible.

Question 8. Consider the problem

maximize 2x1 + 3x2

x1 , x2
subject to 2x1 + x2 ≤ 5,
x1 + x2 ≤ 3,
x1 ≥ 0,
x2 ≥ 0

The optimal solution of the problem is (0, 3). What is the optimal solution of the dual problem? Use
strong duality and complementary slackness.
(a) (1.8, 0) (b) (0, 3) (c) (1, 34 ) (d) (0, 0)

Solution: (b) (0, 3)

Explanation: The dual problem is

minimize 5y1 + 3y2

y1 , y2
subject to 2y1 + y2 ≥ 2,
y1 + y2 ≥ 3,
y1 ≥ 0,
y2 ≥ 0

The objective value of the dual problem is 5y1 + 3y2 . Note that the constraint 2x1 + x2 ≤ 5 does
not hold with equality for (0, 3). By complementary slackness the corresponding dual multiplier y1
is zero. By strong duality, the optimal objective value is 9. As y1 = 0 we have y2 = 3.

2
Question 9. Consider the problem

maximize 2x1 + x2
x1 , x2
subject to x1 + x2 ≤ 4,
x1 − x2 ≤ 2,
x1 ≥ 0,
x2 ≥ 0

The dual of above problem has the form

minimize 4y1 + αy2

y1 , y2
subject to y1 + y2 ψ β,
y1 − y2 ψ 1,
y1 ≥ 0,
y2 ≥ 0

Where α, β are integers and ψ ∈ {≥, ≤, =}. What are the values of α, β, ψ?
(a) (α, β, ψ) = (2, 1, =) (b) (α, β, ψ) = (2, 2, ≥) (c) (α, β, ψ) = (1, 2, ≥) (d) (α, β, ψ) = (2, 2, ≤)

Solution: (b) (α, β, ψ) = (2, 2, ≥)

Explanation: Discussed in lecture during complementary slackness

Question 10. Consider the primal and dual problem in the previous exercise (question 9). Which of
the following is primal and dual optimal solution.
(a) Primal solution (2, 2), dual solution (1, 1)
(b) Primal solution (3, 1), dual solution (1.5, 0.5)
(c) Primal solution (4, 2), dual solution (2, 1)
(d) Primal solution (2, 1), dual solution (1, 0.5)
(e) None of the above

Solution: (b) Primal solution (3, 1), dual solution (1.5, 0.5)
Explanation: (3, 1) is primal feasible, (1.5, 0.5) is dual feasible and complementary slackness is
satisfied as all the constrained are tight.

Question 11. State TRUE or FALSE: The dual of the dual problem is the primal problem.
(a) TRUE (b) FALSE

Solution: (a) TRUE

Explanation: Discussed in lecture.

Question 12. Consider the following network with source s and sink t.

For the min-cut problem the optimal cut is produced by the partition {s, a, c, d} and {b, t}. What is
the maximum flow from source to sink?
(a) 14 (b) 23 (c) 43 (d) 34

3
Solution: (b) 23.
Explanation: Max-flow is the dual of Min-cut problem. By strong duality optimal value will be
same. The optimal cut for min-cut is given and the value of the cut is 12 + 7 + 4 = 23