Genetic Algorithms

• The basic idea of a GA is that the agents derive their properties from a gene sequence.
• That sequence can be a set of binary bits, numbers or something else.
• For example, a solution to the travelling salesman problem with 25 cities might be a gene with 25 elements each being a letter between A and Y.
• Initially, you start with a population of agents.
• After each cycle, they are evaluated.
• The ones that evaluate better reproduce.
• Reproduction can involve crossing two good agents, and mutation.
• Other operators can be used, and there are issues about leaving old members in the population, randomness, uniqueness and other things.
• You need to run these things for several generations.