Skip to main content

Quality-Diversity algorithms in Python

Project description

Package qdpy implements recent Quality-Diversity algorithms: Map-Elites, CVT-Map-Elites, NSLC, SAIL, etc. QD algorithms can be accessed directly, but qdpy also includes building blocks that can be easily assembled together to build your own QD algorithms. It can be used with parallelism mechanisms and in distributed environments.

This package requires Python 3.6+.

qdpy includes the following features:
  • Generic support for diverse Containers: Grids, Novelty-Archives, Populations, etc

  • Optimisation algorithms for QD: random search methods, quasi-random methods, evolutionary algorithms

  • Support for multi-objective optimisation methods

  • Possible to use optimisation methods not designed for QD, such as [CMA-ES](https://arxiv.org/pdf/1604.00772.pdf)

  • Parallelisation of evaluations, using parallelism libraries, such as multiprocessing, concurrent.futures or [SCOOP](https://github.com/soravux/scoop)

  • Easy integration with the popular [DEAP](https://github.com/DEAP/deap) evolutionary computation framework

Install

qdpy requires Python 3.6+. It can be installed with:

pip3 install qdpy

qdpy includes optional features that need extra packages to be installed:
  • cma for CMA-ES support

  • deap to integrate with the DEAP library

  • tables to output results files in the HDF5 format

  • tqdm to display a progress bar showing optimisation progress

  • colorama to add colours to pretty-printed outputs

You can install qdpy and all of these optional dependencies with:

pip3 install qdpy[all]

The latest version can be installed from the GitLab repository:

pip3 install git+https://gitlab.com/leo.cazenille/qdpy.git@master

Example

From a python shell:

from qdpy import algorithms, containers, benchmarks, plots

# Create container and algorithm. Here we use MAP-Elites, by illuminating a Grid container by evolution.
grid = containers.Grid(shape=(64,64), max_items_per_bin=1, fitness_domain=((0., 1.),), features_domain=((0., 1.), (0., 1.)))
algo = algorithms.RandomSearchMutPolyBounded(grid, budget=60000, batch_size=500,
        dimension=3, optimisation_task="maximisation")

# Create a logger to pretty-print everything and generate output data files
logger = algorithms.AlgorithmLogger(algo)

# Define evaluation function
eval_fn = algorithms.partial(benchmarks.illumination_rastrigin_normalised,
        nb_features = len(grid.shape))

# Run illumination process !
best = algo.optimise(eval_fn)

# Print results info
print(algo.summary())

# Plot the results
plots.default_plots_grid(logger)

print("All results are available in the '%s' pickle file." % logger.final_filename)

Usage, Documentation

Please to go the GitLab repository main page (https://gitlab.com/leo.cazenille/qdpy) and the documentation main page (https://leo.cazenille.gitlab.io/qdpy/).

Author:

Leo Cazenille, 2018-*

License:

LGPLv3, see LICENSE file.

Project details


Download files

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

Source Distribution

qdpy-0.1.2.tar.gz (777.2 kB view details)

Uploaded Source

Built Distribution

qdpy-0.1.2-py3-none-any.whl (825.5 kB view details)

Uploaded Python 3

File details

Details for the file qdpy-0.1.2.tar.gz.

File metadata

  • Download URL: qdpy-0.1.2.tar.gz
  • Upload date:
  • Size: 777.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/2.0.0 pkginfo/1.5.0.1 requests/2.24.0 setuptools/47.3.1 requests-toolbelt/0.9.1 tqdm/4.46.1 CPython/3.8.3

File hashes

Hashes for qdpy-0.1.2.tar.gz
Algorithm Hash digest
SHA256 86b9b5669b7fbcbd9c0cfe98c4dac83df940754495dda2062c84d09e27ae70f9
MD5 acaf9751b65d51f6145da3d180c93d70
BLAKE2b-256 522d5bf6bd54bf0ed9722f9a60182737029815d376a6b407fe0a5b716a9fb4db

See more details on using hashes here.

File details

Details for the file qdpy-0.1.2-py3-none-any.whl.

File metadata

  • Download URL: qdpy-0.1.2-py3-none-any.whl
  • Upload date:
  • Size: 825.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/2.0.0 pkginfo/1.5.0.1 requests/2.24.0 setuptools/47.3.1 requests-toolbelt/0.9.1 tqdm/4.46.1 CPython/3.8.3

File hashes

Hashes for qdpy-0.1.2-py3-none-any.whl
Algorithm Hash digest
SHA256 d16039a2b7c74f217c8d35e38511898af1954bef1b82ed43f93a6b3219bfd744
MD5 6cd0a8c768c16f7cc8e0b364395fac34
BLAKE2b-256 85313a3a6b6552c4535bace530a931bdb91dedaceb0e2157418fb79a94adddaf

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page