SimonBlanke
Implement genetic algorithm
Hello @igaloly,
thank you for the suggestion.
I looked into the package you posted. It turns out, that DEAP is the basis for Tpot, which inspired me to create my own package for optimization.
Some of the algorithms in the DEAP package are not fit to be implemented in GFO (because they require a continuous search-space). But I will look into more of the algorithms within the next weeks. If you have a suggestion for a particular algorithm you can post it here and explain it and why it fits GFO.
After doing some research I further understood the difference between evolution strategies and genetic algorithms. I would like to add a genetic-algorithm-optimizer to GFO and updated the title of this issue accordingly.
I created an implementation of the genetic-algorithm-optimizer. I will do some further development in the next days, finish the parameter API and add automatic tests. This is how a search-path-gif of the algorithm solving a rastrigin-function looks like:
I also added the differential evolution algorithm to the sample branch. Also see SimonBlanke/Gradient-Free-Optimizers#48. Now I will refactor the new code. Since GFO has 3 different evolutionary algorithms I would like to use a EvolutionaryAlgorithm-superclass for common-methods and attributes.