Ones counting with callbacks

In this example, we introduce the Callback instances. When creating a new instance of a GeneticAlgorithm, the user has the option of providing a list of callbacks, objects that are called whenever an event is triggered. The previous Ones counting examples will be used as base for this one.

There is a class in the library called Callback, which defines the possible events, but actually any object with the required methods is valid as a callback. The methods, corresponding to the events, are self-explanatory, and are the following:

• on_algorithm_begins(self, ga: api.EvolutiveAlgorithm)
• on_algorithm_ends(self, ga: api.EvolutiveAlgorithm)
• on_step_begins(self, ga: api.EvolutiveAlgorithm)
• on_step_ends(self, ga: api.EvolutiveAlgorithm)

Defining a callback

Our example defines a callback called MyCallback which inherits from the Callback helper class, redefining the four methods:

class MyCallback(Callback):
   def on_algorithm_begins(self, g):
       print('Start algorithm')

   def on_step_begins(self, g):
       print(f'Generation: {g.generation}\t', end='')

   def on_step_ends(self, g):
       print(f'{g.best().phenotype()}\tfitness: {g.best().fitness():.2f}')

   def on_algorithm_ends(self, g):
       print('End algorithm')

Algorithm configuration

The configuration is the same as the previous example with the difference that we’ve added a new argument, callback with a list of objects to be used as callbacks; in our case, only one, an instance of MyCallback:

ga = GeneticAlgorithm(
    # ...
    # Stuff
    # ...
    callbacks=[MyCallback()],
)

Then, during the execution of the algorithm, each time an event is triggered (e.g. a new algorithm iteration or step begins), the corresponding methods from each callback in the list (e.g. on_step_begins) are called.