Next: The genetic operators for
Up: The genetic algorithm
Previous: Selection by deletion of
Contents
Index
Evaluator for operators
To determine which operator to selected next, they are evaluated. Operators, which are rated better in a situation, have a higher probability of being selected respectively run in a similar situation.
The situation may include:
- how many operations wher executed.
- of which nature the original multimedia object is (with the help of the domains for the environment in the root-elements of it):
- it contains colors
- it is monochrome
- it contains sound
- it is a film
- ...
- the average (relative) fitness of individuals.
- the fitness of the best/worst individual.
- the standard deviation of fitness values in the population.
- the number of individuals.
- the operations that have been executed.
- ...
The different evaluation algorithms can be selected. Thus, different evaluation algorithms can be easily interchanged and compared.
To evaluate the operators, data of their previous execution may be kept permanently (by the seperate evaluators). In this way, the algorithm can learn from previous operator calls.
The evaluation of the operators should be system independent, ie. independent of the computer, on which the algorithm is running.
The evaluation criteria can be:
- execution time of the operation
- reached deterioration or improvement
- reliability of the operator (Is always a result returned? Did it crashes sometimes?)
- ...
Next: The genetic operators for
Up: The genetic algorithm
Previous: Selection by deletion of
Contents
Index
Betti Österholz
2013-02-13