GenesPy

A pure python genetic algorithms library. The intention is to create a library without any external dependence. Standard Python only.

Usage guide for a simple case

1. Create a Task object.

   task = Task()
   

2. Write your own evaluation function. This function must accept two parameters:

   • A list (the genome). Each item in the list is a gene.
   • An arbitrary object (which is defined by you, as complementary data to evaluate the individuals). By default, this object is None.

   The function must return a floating value as fitness of the individual.

   def my_eval_function(genome, data):
       # x - genome[0]
       # y - genome[1]
   
       return (genome[0] * data[0]) + (genome[1] * data[1])
   

3. Define the structure of your variables. These can be floating values, binary strings, or permutations:

   # Declaration of binary string structure as tuple of tuples (2 variables)
   struct = ((True, 5, 5),  # x (sign bit, integer bits, mantissa bits)
            (True, 5, 5))   # y (sign bit, integer bits, mantissa bits)
   

4. Create a population and assign to the Task object.

   the_pop = init_binary_pop(1000, struct)  # Binary population
   data = [5, 20]  # Arbitrary data
   task.set_population(the_pop)
   task.set_data(data)
   

5. Choose the selection, crossover and mutation functions that will be used, and assign them to the Task object as well.

   task.set_evals([my_eval_function], [-1.0])  # (-1 minimizes, +1 maximizes)
   task.set_mutator(mutate_flip, {'mp': .3})
   task.set_crossover(crossover_one_point)
   task.set_selector(select_vasconcelos, {'cp': .5})
   

6. Define the desired execution parameters (generations, execution times, etc.). Choose a genetic algorithm and ...

7. Run it!

   elitism = 1.0  # % of puplation in elitism
   duration = 10  # Max duration (in seconds)
   gens = 1000  # Maximum number of generations
   verbose = 2  # Report every n generations
   answer = general_ga(
       task,
       elitism,
       duration,
       gens,
       verbose
   )
   

More examples

Check the examples directory to see the solution of simple problems, using different ways to encode the genome.

TODO

• Documentation
• More selection algorithms
• Add multiobjective selector(s)
• More mutation operators
• More crossover operators
• User documentation and examples