Genetic Programming Flashcards
What is symbolic regression?
Creates a function that matches the values(input and output values)
collection of I/O values = ?
fitness cases
What does regression do?
seek to predict a numeric outcome for a given input
What is the only real access the system has to the problem?
the fitness function
Can we use a function that tell us how many of the fitness cases the program got right?
Yes but it isnt the most informative
What would be the ideal fitness function?
Depends on specific characteristics
Since both the fitness function + fitness cases are numeric what can we do?
Use a function that sums up the difference between the actual value and the expected value for the fitness cases.
What does the overall fitness represent?
The error margin for all fitness cases
How does the GP algorithm look like?
Create init pop of programs
Execute each program and establish fitness
while( termination cond not met){
select fitter programs to reproduce
Create new programs using genetic operators
and update pop
execute each new program and establish the fitness
}
return best program
What kind of structure is a GP?
A tree
What are the tree generation methods?
Full
Grow
Ramped-half and half
What are the initial population parameters?
Initial tree depth
Max tree depth
Pop size
What is important about the fitness function?
Must be effective measure of the goodness of the program to solve the problem
What are the 2 selection methods of a GP
Tournament selection
Fitness proportionate
Read up on subtree cross over and Mutation
