

Functional API

AutoGOAL's functional API allows you to transform any Python callable (e.g., a method) into an optimizable target. In contrast with the class-based and graph-based APIs, the functional API does not require to know before-hand the structure of the space you want to optimize.

This enables very flexible use cases, in which you can iterate quickly, experiment, and transform deterministic code to solve one particular task into optimizable software that seems to magically solve the problem for you in the best possible way.

Let's start with a toy example just to show the basic usage of the API.

from autogoal.sampling import Sampler


def generate(sampler: Sampler):
    x1 = sampler.continuous(0, 1, "x1")
    x2 = sampler.continuous(0, 1, "x2")

    if x1 > x2:
        return (x1, x2)

    return (0, 0)


def fn(t):
    x1, x2 = t
    return x1 * x2


search = PESearch(generate, fn)
best, y = search.run(1000)

print(search._model)

print(best, y)