chapter 4 Flashcards
what is a representation?
- genotype space
- mapping
>>> often representation is used to refer to the genotype space
what is representation discontinuity?
representation discontinuity:
> difference in phenotypic vs genotypic distance
> e.g. phenotype = positive ints, genotype = bit strings
what is the strong duality principle?
strong duality:
> small changes in underlying structuve of an object should cause small changes in the objects behavior
> crucial concept in EC
binary representation: how does 1 point crossover work?
1 point crossover
- choose random point on the two parents
- split parents at this point
- create children by exchanging tails
binary representation: whats a problem with one point crossover?
positional bias
> keeps genes together that are close to each other
> performance depends on the order the variables occur in the representation
binary representation: how does n point crossover work?
n point crossover
- choose n random crossover points
- split along those points
- glue parts together alternating between parents
>>> still has some amount of positional bias
binary representation: how does uniform crossover work?
uniform crossover:
- flip coin for each gene of the first child
- inverse selection for the second child
>>> inheritance is independent of position
crossover is …?
mutation is …?
crossover is explorative
mutation is exploitative
real value representation: how does uniform mutation work?
uniform mutation:
- x’ drawn randomly from [lowerbound, upperbound]
>>> analoguous to bit flipping
real value representation: how does non-uniform crossover work?
> whats the step size
non- uniform crossover:
- most common: add random noise to each var separately from gauss
>>> std is step size
real value representation: how does self adaptive mutation work?
self adaptive mutation:
> step sizes are included in the genome and undergo variation and selection as well
> step is not size by user but co-evolves
why is order of mutation important?
- first mutate sigma, then x
- x’is good is f(x’) is good
- sigma’ is good if the x’ it created is good
>>> other way round does not work
real valued recombination: two types of recombination
- discrete: each allele value comes from one of its parents with equal prob
- intermediate aka arithmetic crossover
> z = alpha*x + (1-alpha)*y
arithmetic crossover 3 possible choices for alpha
- constant > uniform arithmetical crossover
- variable > depending on age of the population
- random
real valued representation: how does simple arithmetic crossover work
simple arithmetic crossover:
> pick random gene, after this point mix values
