7
Genetic Programm ing The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed by the algorithm. In each generation there is a complete population of individuals. The new population is created from the old population which it replaces by applying genetic operators to the individuals. The size of the population remains constant throughout a run of the genetic programming system.

The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Embed Size (px)

Citation preview

Page 1: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic Programming

The Generational Control Model

This is the control model that is traditionally used by GP systems.

There are a distinct number of generations performed by the algorithm.

In each generation there is a complete population of individuals.

The new population is created from the old population which it replaces by applying genetic operators to the individuals.

The size of the population remains constant throughout a run of the genetic programming system.

Page 2: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic ProgrammingGenerational Algorithm

Create the initial population Repeat

Evaluate the population Select parents for the next generation Apply genetic operators

Until the termination criteria are met

Page 3: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic Programming

Generational Control Model Example

Generation 0

c%

c b

*

a dd

+

c *

c

*

c c

+

-

cb cc

a

+

-*

c

*

e f c% ca

c b

Generation 1

Mutate

+

-*

c

*

e f cab

Reproduce

*

a dd

Mutate

c%

b*

c

*

a a

Crossover

c

+

a

+

*

c

*

c c

-

cb cc

Page 4: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic ProgrammingSteady-State Control Model

This model does not implement a fixed number of generations.

A single population of a fixed size is maintained during a genetic programming run.

Newly created individuals replace the individuals with poor fitness in the population.

Inverse selection methods are used for choosing which members of the population to replace. Inverse fitness proportionate selection Inverse tournament selection

Page 5: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic Programming

Steady-State Example Algorithm

Create an initial population. Evaluate each member of the population Use a selection method to select members to apply the

genetic operators to. These are referred to as winners. Apply the genetic operators to the winners. Use inverse selection methods to select members of the

population to replace with the newly created offspring. Replace the losers in the population with the offspring. Repeat the second through to the seventh step until the

termination criterion is met.

Page 6: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic Programming

Steady-State Control Model Example

Population

c%

c b

*

a dd

+

c *

c

*

c c

+

-

cb cc

a

+

-*

c

*

e f c% ca

c b

17.25 23.76

10.1133.32

15.33

Updated Population

mutate

%

a b

*

dd

*

a dd

+

-

cb cc

a

+

-*

c

*

e f c% ca

c b

replace

+

c *

c

*

c c

Page 7: The Generational Control Model This is the control model that is traditionally used by GP systems. There are a distinct number of generations performed

Genetic Programming

Varying Population Size Control

In this case a single population is maintained, the size of which changes throughout a GP run.

This algorithm enables the population size to grow and shrink according to the state of the search.

If the fitness is high then the population grows. If the neighbourhood of an optimal solution is

found the population size is decreased. In the neighbourhood of a local optimum the

size of the population is increased.