[sage-hamburg] Genetic Algorithms

[Micha Beyer]

Hallo Leute,

am 2.7.2007 ab ca. 19.30Uhr wird Rune folgender Vortrag im Barmbek bei 
Balista halten.

###########################################
Genetic Algorithms

In the end of 2005 I wrote my bachelor project about genetic
algorithms and neural networks.
Not so much to solve an actual problem, as to take a close look at
genetic algorithms and see how you can use them.
In this little presentation I first give a very short introduction to
artificial intelligence, learning algorithms and evolutionary
algorithms. Then an in depth work through of the genetic algorithm
paradigm, where it will be explained how they work and how you can
program them.

Genetic algorithms are used to find a good solution to optimization
problems. They are inspired by evolutionary biology, and use methods
such as selection (survival of the fittest), inheritance (genome),
mutation and crossover (reproduce).

Here is a pseudo-code algorithm:
1. Create the first population.
2. Do loop:
    1. Evaluate population to find fitness of the individuals
    2. Select individuals suited for reproduction.
    3. Create new individuals by crossover
    4. Mutate individuals.
    5. Replace old population with new population.
3. Until stop condition is reached.

It is in fact a simple algorithm. But it can leave you with a lot of
questions:
How do you represent an individual?
How do you determine an individual's fitness?
How do you make a crossover?
How much mutation is necessary?
Etc.

The short answer is, it is problem dependent, - but there seems to be
some general answers anyway, and  this is what I hope to tell you more
about.
##########################################

http://www.lug-balista.de
-- 
Mfg,
 Michael