You are on page 1of 3

ISyE 6669 Homework 15

Problem 1
(a)
Let 𝑥𝑡 = 1 if there is a transaction in period 𝑡 and 0 if not. If there is a transaction in the
first period, then we can have transactions in at most half of the remaining periods.
𝑇
1
∑ 𝑥𝑡 ≤ 𝑥1 𝑇 + 𝑀(1 − 𝑥1 )
2
𝑡=2

If 𝑥1 = 1 (i.e., there is a transaction in the first period) then the second term on the right-
1
hand side of the inequality will go away and the first term will equal 2 𝑇. This will force
the sum of 𝑥2 , … , 𝑥𝑇 to be less than or equal to half of T, which means we can have
transactions in at most half of the remaining periods. M can be set equal to 𝑇 − 1.

(b)
Write a set of constraints to model the requirement that:
|2𝑥1 − 𝑥2 − 𝑥3 | ≥ 2
by introducing an additional binary variable. We can achieve this by introducing a
variable to indicate if 2𝑥1 − 𝑥2 − 𝑥3 > 0 to take advantage of the fact the absolute value
function can be represented by a two-piece linear function depending on the sign of the
function value.
2𝑥1 − 𝑥2 − 𝑥3 + 𝑀𝑦 ≥ 2
−2𝑥1 + 𝑥2 + 𝑥3 + 𝑀(1 − 𝑦) ≥ 2
When 𝑦 = 0, the first constraint becomes 2𝑥1 − 𝑥2 − 𝑥3 ≥ 2, which ensures that the left-
hand side must be positive and greater than 2. When M is set large enough, the second
constraint is always fulfilled. When 𝑦 = 1, it implies that 2𝑥1 − 𝑥2 − 𝑥3 < 0, and the
second constraint becomes active. In this case, M can be set to absolute value of the
minimum possible value 2𝑥1 − 𝑥2 − 𝑥3 − 2 or −2𝑥1 + 𝑥2 + 𝑥3 − 2 can take given that all
variables are in the range [0, 3]. It turns out this minimum is −8, so 𝑀 = 8.

(c)
Write a set of constraints to model the requirement that either 𝑥1 + 𝑥2 ≤ 10 or 2𝑥1 − 𝑥2 ≥
5 but not both.
𝑥1 + 𝑥2 − 𝑀1 𝑦 ≤ 10
𝑥1 + 𝑥2 + 𝑀2 (1 − 𝑦) ≥ 11
2𝑥1 − 𝑥2 − 𝑀3 𝑦 ≤ 6
2𝑥1 − 𝑥2 + 𝑀4 (1 − 𝑦) ≥ 5

This study source was downloaded by 100000843852835 from CourseHero.com on 04-24-2023 21:41:17 GMT -05:00

https://www.coursehero.com/file/123151464/ISyE-6669-Homework-15pdf/
We introduced a binary variable, 𝑦, such that when 𝑦 = 1, the following two constraints
become active:
𝑥1 + 𝑥2 ≥ 11
2𝑥1 − 𝑥2 ≥ 5
which ensures that only the second original constraint is met. When 𝑦 = 0, the following
two constraints become active:
𝑥1 + 𝑥2 ≤ 10
2𝑥1 − 𝑥2 ≤ 6
which ensures that only the first original constraint will be met. We set the big-M
variables as follows to guarantee the constraints will be satisfied when 𝑀𝑖 𝑦 ≠ 0 or
𝑀1 (1 − 𝑦) ≠ 0:
𝑀1 = 10
𝑀2 = 11
𝑀3 = 14
𝑀4 = 15

Problem 2
The objective to minimize the total dose deposited on all normal cells can be written as:
100 5000

min ∑ ∑ 𝑥𝑖 𝑑𝑖𝑗
𝑥
𝑖=1 𝑗=1001

Constraint 1: The dose deposited on each cancerous cell should be at least 70 units.
100

∑ 𝑥𝑖 𝑑𝑖𝑗 ≥ 70 ∀ 𝑗 ∈ 1, … ,1000
𝑖=1

Constraint 2: At least 90% of the cancerous cells should get a dose deposition of at least
75 units.
First, we introduce a set of binary variables 𝑦𝑗 to indicate if cancer cell 𝑗 has a dose
deposition of at least 75 units.
100

∑ 𝑥𝑖 𝑑𝑖𝑗 + 75(1 − 𝑦𝑗 ) ≥ 75 ∀𝑗 ∈ 1, … ,1000


𝑖=1
1000

∑ 𝑦𝑗 ≥ 900
𝑗=1

When 𝑦𝑗 = 1, the total dose given to cancer cell j must be greater than 75. If 𝑦 = 0, the
constraint will still be met, but the total dose does not have to be greater than 75.

This study source was downloaded by 100000843852835 from CourseHero.com on 04-24-2023 21:41:17 GMT -05:00

https://www.coursehero.com/file/123151464/ISyE-6669-Homework-15pdf/
Constraint 4: The dose deposited on each normal cell should not exceed 25 units.
100

∑ 𝑥𝑖 𝑑𝑖𝑗 ≤ 25 ∀ 𝑗 ∈ 1001, … ,5000


𝑖=1

Problem 3
(a)
Given that the objective value, 𝑣0 , is greater than all other objective values down the
tree, this is a maximization problem. This is true because the root node is the original LP
relaxation, and it will always have the best objective value.

(b)
The only node that we still need to branch from is node 9 because it has a fractional
solution and an objective value that above the lower bound of 9.

(c)
We do not need to branch from the following nodes:
• 4 – integer solution
• 5 – integer solution
• 7 – integer solution
• 8 – infeasible
• 10 – objective value is below the lower bound

(d)
The best solution found so far is at node 9, which has an objective value of 9.4. This
solution has an optimality gap of 10 − 9.4 = 0.6.

(e)
A node must have an objective value greater than the current incumbent solution in
order to keep exploring. Therefore, we know that the integer solution at node 7 was
found first, then the integer solution at node 8 and finally the integer solution at node 5.

This study source was downloaded by 100000843852835 from CourseHero.com on 04-24-2023 21:41:17 GMT -05:00

https://www.coursehero.com/file/123151464/ISyE-6669-Homework-15pdf/
Powered by TCPDF (www.tcpdf.org)

You might also like