Giải thuật di truyền với kích cỡ quần thể thay đổi, áp dụng giải bài toán tối ưu đa mục tiêu về khẩu phần thức ăn gia súc

Nguyễn Thu Huyền và Đtg Tạp chí KHOA HỌC & CÔNG NGHỆ 93(05): 75 - 79

75

GENETIC ALGORITHMS VARYING SIZE OF POPULATION

AND MULTI -OBJECTIVE OPTIMIZATION PROBLEMS,

APPLY TO SOLVE ANIMAL FEED OPTIMIZATION PROBLEM

Nguyễn Thu Huyền

1

, Lương Sỹ Ước

2

, Vũ Mạnh Xuân3*

1College of Information and Communication Technology - TNU

2College of Technology and Economics – TNU, 3College of Education - TNU

ABSTRACT

Population size is an important parameter in Genetic Algorithms (GAs). How population size is

reasonable is a matter of concern when designing programs using GAs. Overall, population size is

defined as a given parameters and unchanged in evolution processes. This paper presents research

results that GAs population size changes affecting the diversity of populations and apply to multi￾objective optimization problems, specific the animal feed optimization problem.

Keywords: genetic algorithms, population size, diversity of populations, animal feed optimization

problem.

INTRODUCTION*

In GAs, population size is an important

parameter and is determined after establishing

the program. Population size is how much is

appropriate, it depends on the problem that

we solve. Typically, population size is fixed

mean number of individuals in the population

is unchanged from generation to generation.

But in nature, population size is not fixed, so

the study of GA is applied to the course. First,

it is necessary to determine the initial

population size, this number can change over

generations. But changed will be what? when

we increase the size? when we will increase

only?... are the problems to be solved.

This paper studies and proposes an algorithm

that GAs population size is not fixed. Test

results are presented in multi-objective

optimization problems, special the animal

feed optimization problem.

The paper is structured as follows: After the

preamble is proposed GAs with population

size adjustment. The next section presents

briefly the problem multi-objective

optimization problem and in particular animal

feed optimization problem. Next, a test result

as tools for solving by the proposed algorithm

is compared with traditional algorithm.

*

Tel: 0912 700396

PROPOSE ALGORITHM

For comparison, we present some real-coded

GAs with population size fixed and the

proposed algorithm in which the population

size changes during evolution, as follows:

Traditional algorithm

- Encoding: Each individual is a vector of n

elements; each component is a real number in

the domain of the problem. Thus, a

population with m individuals can be

considered as a real matrix m x n level.

The genetic operators:

- Crossover: Using a crossover point to

generate new individual.

- Mutation is a used mutation.

- Selection: Select the individual to the next

generation by selective competition.

The parameters of the algorithm: population

size fixed (i.e. 50 individuals), crossover

probability is 1, and mutation probability is 0.1.

Proposed algorithm

For comparison, the proposed algorithm is

designed similar to algorithm with population

size fixed, the different of proposed algorithm

and traditional algorithm is just population

size parameter. Once initialized, the

population size is 50 like the original

algorithm above, but from the 2nd generation

onwards we calculate the average adaptation

of individuals with the best adapt for

Số hóa bởi Trung tâm Học liệu – Đại học Thái Nguyên http://www.lrc-tnu.edu.vn