bingo-nasa 0.2.2

A package for genetic optimization and symbolic regression

master:

develop:

General

Bingo is an open source package for performing symbolic regression, Though it can be used as a general purpose evolutionary optimization package.

Key Features

• Integrated local optimization strategies
• Parallel island evolution strategy implemented with mpi4py
• Coevolution of fitness predictors

Note

At this point, the API is still in a state of flux. The current release has a much more stable API but still lacks some of the features of older releases.

Getting Started

Dependencies

Bingo is intended for use with Python 3.x. Bingo requires installation of a few dependencies which are relatively common for data science work in python:

• numpy
• scipy
• matplotlib
• mpi4py (if parallel implementations are to be run)
• pytest, pytest-mock (if the testing suite is to be run)

A requirements.txt file is included for easy installation of dependencies with pip or conda.

Installation with pip:

pip install -r requirements.txt

Installation with conda:

conda install --yes --file requirements.txt

BingoCpp

A section of bingo is written in c++ for increased performance. In order to take advantage of this capability, the code must be compiled. See the documentation in the bingocpp submodule for more information.

Note that bingo can be run without the bingocpp portion, it will just have lower performance.

If bingocpp has been properly installed, the following command should run without error.

python -c "import bingocpp"

A common error in the installation of bingocpp is that it must be built with the same version of python that will run your bingo scripts. The easiest way to ensure consistent python versioning is to build and run in a Python 3 virtual environment.

Documentation

Sphynx is used for automatically generating API documentation for bingo. The most recent build of the documentation can be found in the repository at: doc/_build/html/index.html

Running Tests

An extensive unit test suite is included with bingo to help ensure proper installation. The tests can be run using pytest on the tests directory, e.g., by running:

pytest tests 

from the root directory of the repository.

Usage Examples

In addition to the example shown here, the best place to get started in bingo is by going through the examples directory. It contains several scripts and jupyter notebooks.

A simple evolutionary analysis with bingo: the one max problem

This example walks through the general steps needed to set up and run a bingo analysis. The example problem described here is the one max problem. In the one max problem individuals in a population are defined by a chromosome with a list of 0 or 1 values, e.g., [0, 1, 1, 0, 1]. The goal of the optimization is to evolve toward an optimum list containing all 1's. A complete version of this example is script is found here.

Defining a chromosome generator

Bingo's built-in MultipleValueChromosome is used here. Individuals of this contain their genetic information in a list attribute named values. A chromosome generator is used to generate members of the population. The MultipleValueChromosomeGenerator generates these individuals by populating the indivudual's values from a given input function.

import numpy as np
from bingo.Base.MultipleValues import MultipleValueChromosomeGenerator
np.random.seed(0)  # seeded for reproducible results

def generate_0_or_1():
    return np.random.choice([0, 1])

generator = MultipleValueChromosomeGenerator(generate_0_or_1,
                                             values_per_chromosome=16) 

Defining the evolutionary algorithm

Evolutionary algorithms have 3 phases in bingo: variation, evaluation and selection. The variation phase is responsible for generating offspring of the population, usually through some combination of mutation and crossover. In this example VarOr is used which creates offspring through either mutation or crossover (never both).

from bingo.Base.MultipleValues import SinglePointCrossover, SinglePointMutation
from bingo.Base.VarOr import VarOr

crossover = SinglePointCrossover()
mutation = SinglePointMutation(generate_0_or_1)
variation_phase = VarOr(crossover, mutation,
                        crossover_probability=0.4,
                        mutation_probability=0.4)

The evaluation phase is responsible for evaluating the fitness of new members of a population. It relies on the definition of a FitnessFunction class.
The goal of bingo analyses is to minimize fitness, so fitness functions should be constructed accordingly. In the one max problem fitness is defined as the number of 0's in the individuals values.

from bingo.Base.FitnessFunction import FitnessFunction
from bingo.Base.Evaluation import Evaluation

class OneMaxFitnessFunction(FitnessFunction):
    """Callable class to calculate fitness"""
    def __call__(self, individual):
        return individual.values.count(0)

fitness = OneMaxFitnessFunction()
evaluation_phase = Evaluation(fitness)

The selection phase is responsible for choosing which members of the population proceed to the next generation. An implementation of the common tournament selection algorithm is used here.

from bingo.Base.TournamentSelection import Tournament

selection_phase = Tournament(tournament_size=2)

Based on these phases, an EvolutionaryAlgorithm can be made.

from bingo.Base.EvolutionaryAlgorithm import EvolutionaryAlgorithm

ev_alg = EvolutionaryAlgorithm(variation_phase, evaluation_phase, 
                               selection_phase)

Creating an island and running the analysis

An Island is the fundamental unit in bingo evolutionary analyses. It is responsible for generating and evolving a population (using a generator and evolutionary algorithm).

from bingo.Base.Island import Island

island = Island(ev_alg, generator, population_size=10)
best_individual = island.best_individual()
print("Best individual at start: ", best_individual)
print("Best individual's fitness: ", best_individual.fitness)

    Best individual at start:  [1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1]
    Best individual's fitness:  5

The island can be evolved directly using it's execute_generational_step member function. In this case the population is evolved for 50 generations

for _ in range(50):
    island.execute_generational_step()

best_individual = island.best_individual()
print("Best individual at end: ", best_individual)
print("Best individual's fitness: ", best_individual.fitness)

    Best individual at end:  [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
    Best individual's fitness:  0

Contributing

1. Fork it (https://github.com/nasa/bingo/fork)
2. Create your feature branch (git checkout -b feature/fooBar)
3. Commit your changes (git commit -am 'Add some fooBar')
4. Push to the branch (git push origin feature/fooBar)
5. Create a new Pull Request

Versioning

We use SemVer for versioning. For the versions available, see the tags on this repository.

Authors

• Geoffrey Bomarito
• Kathryn Esham
• Ethan Adams
• Tyler Townsend
• Diana Vera

License

Notices

Copyright 2018 United States Government as represented by the Administrator of the National Aeronautics and Space Administration. No copyright is claimed in the United States under Title 17, U.S. Code. All Other Rights Reserved.

The Bingo Mini-app framework is licensed under the Apache License, Version 2.0 (the "License"); you may not use this application except in compliance with the License. You may obtain a copy of the License at [http://www.apache.org/licenses/LICENSE-2.0].

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.