Page 175 - Data Science Algorithms in a Week

The Estimation of Cutting Forces in the Turning of Inconel 718 Assisted … 159

permutation encoding, value encoding and tree encoding are also used as an encoding
methods in genetic algorithms.
Creation of the initial population. The GA sequence begins with the creation of an
initial population of individuals. The most common way to do this is to generate a
population of random solutions. A population of individuals represents a candidate
solution to the problem and the population size depend on the complexity of the problem.
Ideally, the first population should have a gene pool as large as possible in order to
explore the whole search space. Nevertheless, sometimes a problem specific knowledge
can be used to construct the initial population. Using a specific heuristic to construct the
population may help GA to find good solutions faster, but the gene pool should be still
large enough. Furthermore, it is necessary to take into account the size of the population.
The larger population enable easier exploration of the search space, but at the same time
increases the time required by a GA to converge.
Selection. Selection is process of randomly picking chromosomes out of the
population according to their evaluation function, where the best chromosomes from the
initial population are selected to continue, and the rest are discarded. The members of the
population are selected for reproduction or update through a fitness-based process, where
the higher the fitness function, guarantee more chance for individual to be selected. The
problem is how to select these chromosomes and there are a large number of methods of
selection which have been developed that vary in complexity. A method with low
selectivity accepts a large number of solutions which result in too slow evolution, while
high selectivity will allow a few or even one to dominate, which result in reduction of the
diversity needed for change and progress. Therefore, balance is needed in order to try
prevent the solution from becoming trapped in a local minimum. Several techniques for
GA selection have been used: the roulette wheel, tournament, elitism, random, rank and
stochastic universal sampling,
Reproduction. Reproduction is the genetic operator used to produce new generation
of populations from those selected through selection using two basic types of operators:
crossover and mutation. Crossover operators selects genes from parent chromosomes and
creates a new offspring. The simplest way to do this is to choose any crossover point on
the string and everything after and before the point is crossed between the parents and
copied. There are several types of crossover operators: single-point crossover, two-point
crossover, multi-point crossover, uniform crossover, three parent crossover, crossover
with reduced surrogate, shuffle crossover, precedence preservative crossover, ordered
crossover and partially matched crossover. The basic parameter of crossover operator is
the crossover probability which describe how often crossover will be performed. If
crossover probability is 0%, then whole new generation is made from exact copies of
chromosomes from old population, elsewise if it is 100%, then all offspring are made by
crossover. After crossover, the mutation operator is applied on the strings. Mutation
ensure more variety of strings and prevent GA from trapping in a local minimum. If task