Skip to main content
Join the official 2020 Python Developers SurveyStart the survey!

A command line tool for running experiments with `edo`.

Project description


PyPI version CI Code style: black DOI

A command line tool for running experiments with edo.


edolab is pip-installable:

$ python -m pip install edolab


Experiment scripts

To use edolab, you will need to write a Python script configuring the parameters of your experiment.

Required parameters

  • fitness: A function that takes (at least) an edo.Individual instance to be used as the fitness function by edo
  • distributions: A list of edo.distribution.Distribution subclasses that will be used to create the edo.Family instances for edo
  • Variable assignments for all of the essential arguments in edo.DataOptimiser except for families

Optional parameters

  • root: A directory to which data should be written (and summarised from)
  • processes: A number of processes for edo to use when calculating population fitness
  • Custom column distribution classes should be defined in the script
  • If you wish to use a custom stop or dwindle method then define a subclass of edo.DataOptimiser and assign that class to a variable called optimiser
  • Any keyword arguments to pass to fitness or the stop and dwindle methods should be assigned to the corresponding <func>_kwargs variable.

An example of such a script would be something like this:

""" /path/to/experiment/ """

import edo
import numpy as np
from edo.distributions import Uniform

class CustomOptimiser(edo.DataOptimiser):
    """ This is an optimiser with custom stopping and dwindling methods. """

    def stop(self, tol):
        """ Stop if the median fitness is less than `tol` away from zero. """

        self.converged = abs(np.median(self.pop_fitness)) < tol

    def dwindle(self, rate):
        """ Cut the mutation probability in half every `rate` generations. """

        if self.generation % rate == 0:
            self.mutation_prob /= 2

def fitness(individual, size, seed=0):
    """ Randomly sample `size` values from an individual and return the
    minimum. """

    values = individual.dataframe.values.flat
    sample = np.random.choice(values, size=size)
    return min(sample)

class NegativeUniform(Uniform):
    """ A copy that only takes negative values. """

    name = "NegativeUniform"
    param_limits = {"bounds": [-1, 0]}
    hard_limits = {"bounds": [-100, 0]}

size = 5
row_limits = [1, 5]
col_limits = [1, 2]
max_iter = 3
best_prop = 0.5
mutation_prob = 0.5

Uniform.param_limits["bounds"] = [0, 1]

distributions = [Uniform, NegativeUniform]
optimiser = CustomOptimiser

fitness_kwargs = {"size": 3}
stop_kwargs = {"tol": 1e-3}
dwindle_kwargs = {"rate": 10}

For more details on the parameters of edo, see its documentation:

Running the experiment

Then, to run an experiment with this script do the following:

$ edolab run /path/to/experiment/

Summarising the experiment

And to summarise the data (for easy transfer):

$ edolab summarise /path/to/experiment/

For further details on the commands, use the --help flag on the run and summarise commands.

A note on reproducibility

It is highly recommended that you use a virtual environment when using edo in or outside of this command line tool as edo uses pickle to store various objects created in a run that may not be retrievable with a different version of Python.


This tool has been made to be pretty bare and could use some padding out. If you'd like to contribute then make a fork and clone the repository locally:

$ git clone<your-username>/edolab.git

Install the package and replicate the conda environment (or install the development dependencies manually):

$ cd edolab
$ python develop
$ conda env create -f environment.yml
$ conda activate edolab-dev

Make your changes and write tests to go with them, ensuring they pass:

$ python -m pytest --cov=edolab --cov-fail-under=100 tests

Commit, push to your fork and open a pull request!

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for edolab, version 0.0.5
Filename, size File type Python version Upload date Hashes
Filename, size edolab-0.0.5.tar.gz (9.1 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page