Studied by 1 person
evolutionary programs used to represents problems with real value numbers. genetic algorithms used for genes that are binary or a fixed number of permutations
what is evolutionary computation?
family of algorithms/ techniques that are inspired by evolution theory and aimed to solve optimisation problems and generate AI in games
what methods are under the field of evolutionary computation
evolutionary programming
evolutionary strategies
genetic algorithms
genetic programming
learning classifier systems
estimation of distribution algorithms
what is the basic concept of evolutionary computation?
what is evolutionary programming?
multi-population algorithm similar to GAs that aim to find a solution.
unlike GAs, every parent mutates to create a single ‘child’ → this means that the population doubles (instead of using crossover to create a child based on 2 parents)
how does evolutionary programming differ from genetic algrotihms?
mutation operator is more complex and adaptive
does not involve crossover operator
selection operator is in the form of Tournament selection and not a roulette wheel
what is tournament selection?
each member of the population is compared with a fixed number of other individuals
for each comparison the individual gains a point if its fitness is better than the opponent
after comparisons, the population is reduced to its original size (remove 50%) by keeping the individuals with the highest score
what is the pseudocode for the evolutionary programming algorithm?
check in-between tournament selection and mutation to check for invalid individuals
what is genetic programming?
an evolutionary approach that extends genetic algorithms where we evolve computer programs by natural selection
what is one type of genetic programming?
symbolic regression
what is symbolic regression?
a mathematical expression is represented as a tree and we are trying to breed the best mathematical function
how does symbolic regression work?
initially determine the set of terminals and non-terminals/functions
fitness = the observed data vs. the calculated data that comes from evaluating the expression the tree represents
crossover and mutation are adapted to handle tree structures
a genetic programming algorithm is virtually the same as the genetic algorithm
what type of operators can genetic programming do compared to genetic algorithms?
as an extension of genetic algorithms, genetic programming can do all of the operators in GAs plus more specialised versions that work on the mathematical tree rather than on a chromosome (string of digits)
what are the specialised genetic programming operators?
sub-tree exchange crossover
self crossover
point mutation
permutation mutation
prune mutation
hoist mutation
expansion mutation
sub-tree mutation
what is the sub-tree exchange crossover operator?
two sub trees are swapped between parents
what is the self crossover operator?
sub trees are exchanged within an individual parent
what is the point mutation operator?
a node in the tree is changed to a different symbol
what is the permutation mutation operator?
two terminal (num/variable) nodes from the same sub tree are swapped
what is the hoist mutation operator?
a sub tree creates a new individual
what is the expansion mutation operator?
a sub tree is added to the base of the tree
what is the prune mutation operator?
a sub tree is removed
what is the sub tree mutation operator?
a sub tree is replaced with a random sub tree
Note
Flashcard