How to Find Minimum Genetic Algorithms
- 1). Design a string of numbers that describes a solution to your problem. Design a function that can evaluate these strings; given a string, this function can tell you how good the string is as an answer to your problem. For example, the string may be a list of measurements of the interior of a rocket engine turbo pump. The function would be a computation of thrust from a chamber with these measurements. The string could also be the percents of water, sand, cement and gravel that constitute a mixture of undersea concrete.
- 2). Write a computer program that evolves the strings toward optimization. Start with a population of strings where all of the numbers in the strings were chosen at random. A "generation" consists of evaluating all of the strings in the "population" and discarding the strings with the lowest values. The strings with the highest evaluations will continue into the next generation. A random blending of the high preforming strings will also make it to the next generation. Some of these strings will have random mutations to one of the numbers in the string. This process is repeated for several generations. The number of strings in each generation is the same.
- 3). Find the minimum genetic algorithm by running your solution with one of the numbers missing. If there are "n" numbers in a string, this will mean running "n" different populations one with each of the numbers missing. If the shorter strings still have the same high evaluation, run shorter and shorter populations until you find the shortest string that will work for the specific problem. For the concrete problem, this would tell you if you could leave out one of the ingredients.
Source...