ASSIGNMENT

1.Explain the working of NSGA-II. (Flowchart is expected).

NSGA-II is an improved version of the non-dominated sorting genetic algorithm
(NSGA), which has been criticized by researchers due to its limitations such as the
absence of elitism, the need to define sharing parameter for diversity preservation,
and its high computational complexity. On the other hand, the design of NSGA-II
exhibits the property of elitism and does not need any sharing parameter. It uses the
crowding distance operator for the mechanism of diversity preservation. Moreover,
it is computationally fast and lives up to its name ‘Fast Elitist NSGA-II.’ The overall
complexity of NSGA-II is at most O(MN2 ), where M is the number of objective
functions, and N is the population size.
1) BASIC STRUCTURE OF NSGA-II
The philosophy of NSGA-II is based on four main principles, which are: Non-
Dominated Sorting, Elite Preserving Operator, Crowding Distance and Selection
Operator. These are described in brief in the following subsections.
a: NON-DOMINATED SORTING
In this procedure, the population members are sorted using the concept of Pareto
dominance. The process of non-dominated sorting begins with assigning the first
rank to the non-dominated members of the initial population. These first ranked
members are then placed in the first front and removed from the initial population.
After that, the non-dominating sorting procedure is performed on the remaining
population members. Further, the non-dominated members of the remaining
population are assigned the second rank and placed in the second front. This process
continues until the whole population members are put on different fronts according
to their ranks.

b: ELITE-PRESERVING OPERATOR
Elite preserving strategy retains the elite solutions of a population by directly
transferring them to the next generation. In other words, the non-dominated
solutions found in each generation move on to the next generations till some
solutions dominate them.
 c: CROWDING DISTANCE
The crowding distance is calculated to estimate the density of solutions surrounding
a particular solution. It is the average distance of two solutions on either side of the
solution along each of the objectives. On comparing two solutions with different
crowding distances, the solution with the large crowded distance is considered to be
present in a less crowded region. The crowded distance of the ith solution is the
average side-length of the cuboid,). If fij is the jth value of an objective function for
the ith individual and, fjmax and fjmin are the maximum and minimum values.

Non-dominated sorting procedure and Crowding distance calculation.

respectively of jth objective function among all the individuals. Then, the crowding
distance of ith individual is defined as the average distance of two nearest solutions
on either side.
𝑘
𝑓𝑗𝑖+1 − 𝑓𝑗 𝑖−1
𝑐𝑑(𝑖) = ∑
𝑓𝑗 𝑀𝑎𝑥 − 𝑓𝑗 𝑀𝑖𝑛
𝑖=1

where k is the number of objective functions

d: SELECTION OPERATOR The population for the next generation is selected using a
crowded tournament selection operator, which uses the rank of the population
members and their crowding distances for the selection. The rule for selecting one
out of two population members for the next generation is-
(i) If both the population members are of different ranks, then the one with the better
rank is selected for the next generation
(ii) (ii) If both the population members are of the same ranks, then the one with the
higher crowding distance is selected for the next generation.
2) PROCEDURE OF NSGA-II
The procedure of NSGA-II begins with generating an initial population Pt of size N.
Then, a new population Qt is created after performing crossover and mutation
operations on the population Pt. After that, the population Pt and Qt are combined
to form a new population Rt, and the non-dominated sorting procedure is performed
on Rt. Then, the population members of Rt are ranked into different fronts according
to their non-domination levels. The next process is to select N members from Rt to
create the next population Pt+1. If the size of the first front is greater than or equal to
N, then only N members are selected from the least crowded region of the first front
to form Pt+1. On the contrary, if the size of the first front is less than equal to N, then
the members of the first front are directly transferred to the next generation, and
the remaining members are taken from the least crowded region of the second front
and added to Pt+1. If the size of Pt+1 is still less than N, then the same procedure is
followed for the next consecutive fronts until the size of Pt+1 becomes equal to N. The
populations Pt+2, Pt+3, Pt+4, . . ., for the next generations are constructed using the
same procedure until the stopping criteria are not satisfied.

Procedure of NSGA-II
Flowchart of NSGA-II
2. What is the role of Pareto Front?

Let x 1 and x 2 be the two feasible solutions of the Multi objective

minimization problem (1). The solution x 1 can be viewed as better than x 2
if the following conditions hold:
1. fj(x 1 ) ≤ fj(x 2 ) for all j = {1, 2, . . . , k}
2. fj(x 1 ) < fj(x 2 ) for at least one j = {1, 2, . . . , k}
where k is the number of objective functions, fj(x) is the jth value of an
objective function for decision vector x. In this case, we say that x1
dominates x2 (or x2 is dominated by x1 ): x 1 is better than x2 . The relation
‘<’ (or ‘>’ for maximization problem) can be denoted as a dominance
operator ⊲. x1 ⊲ x2 represents x 1 dominates x 2.
When a solution x of is not dominated by any other feasible solutions, it is
called a Pareto optimal solution. The set of all Pareto optimal solutions are
referred to as a Pareto set. The objective vector corresponding to the
Pareto set is defined as a Pareto front.

Pareto dominance.