Learning genetic algorithm and genetic programming
It's a kind of evolutionary algorithm to solve optimization problems.
The GA is composite by:
- Population: What is the size of population?
- Individuals
There are some steps well defined on GA:
1. Initial population: Represents the elements that starts the algorithm. 2. Cross over: The process of two selected parents generated descendants. For instance: P1 and P2 are the parents with the following chromossomes P1 = (0,1,0,0,1,1,0,1) and P2 = (1,1,1,0,0,0,1,0). The crossover could be n-points or uniform. Analyzing a 1-point crossover on 3 chromossome these parents will generate the following offsprings D1 = (0,1,0,0,0,0,1,0) and D2 = (1,1,1,0,1,1,0,1). 3. Mutation: In nature mutation is a kind of error on 'copy' genes and we try to reproduce it on GA. This step ends the reproduce phases. 4. Merge: Merging the parents with the descendants to generate a new population with the right size. The worst members are discarted. 5. Selection: It's the process to select parents to create the next generations of individuals. This process could be done: 5.1. Random: Select the parents randomly is not effective because weak parents could be selected 5.2. Tournament: Selecting few elements (like 5) and do a tournament between them, the winner will be selected 5.3. Roulette wheel: It's a kind of probability selection