wildboar.datasets#

Submodules#

Package Contents#

Classes#

ArffBundle

bundle of .arff-files

NpyBundle

bundle of numpy binary files

Bundle

Base class for handling dataset bundles

Repository

A repository is a collection of bundles

JSONRepository

A repository is a collection of bundles

Functions#

set_cache_dir(cache_dir)

Change the global cache directory

load_synthetic_control([merge_train_test])

Load the Synthetic_Control dataset

load_two_lead_ecg([merge_train_test])

Load the TwoLeadECG dataset

load_gun_point([merge_train_test])

Load the GunPoint dataset

load_datasets([repository, cache_dir, ...])

Load all datasets as a generator

load_all_datasets([repository, cache_dir, ...])

load_dataset(name, *[, repository, dtype, contiguous, ...])

Load a dataset from a repository

list_datasets([repository, cache_dir, ...])

List the datasets in the repository

clear_cache([repository, cache_dir, keep_last_version])

Clear the cache by deleting cached datasets

get_repository(repository)

Get repository by name

install_repository(repository)

Install repository

get_bundles(repository)

Get all bundles in the repository

list_bundles(repository)

Get a list of all bundle names in the specified repository.

list_repositories()

List the key of all installed repositories
class wildboar.datasets.ArffBundle(*, key, version, name, description=None, class_index=-1, encoding='utf-8')#

Bases: Bundle

bundle of .arff-files

class wildboar.datasets.NpyBundle(*, key, version, name, description=None, class_index=-1)#

Bases: Bundle

bundle of numpy binary files

class wildboar.datasets.Bundle(*, key, version, name, description=None, class_index=-1)#

Base class for handling dataset bundles

name#

Human-readable name of the bundle

Type:

str

description#

Description of the bundle

Type:

str

class_index#

Index of the class label(s)

Type:

int or array-like

get_filename(version=None, tag=None, ext=None)#
list(archive)#

List all datasets in this bundle

Parameters:

archive (ZipFile) – The bundle file

Returns:

dataset_names – A sorted list of datasets in the bundle

Return type:

list

load(name, archive, *, dtype=None)#

Load a dataset from the bundle

Parameters:

  • name (str) – Name of the dataset

  • archive (ZipFile) – The zip-file bundle

  • dtype (object, optional) – Cast the data and label matrix to a specific type

Returns:

  • x (ndarray) – Data samples

  • y (ndarray) – Data labels

  • n_training_samples (int) – Number of samples that are for training. The value is <= x.shape[0]

class wildboar.datasets.Repository#

A repository is a collection of bundles

abstract property name#

Name of the repository

Returns:

str

Return type:

the name of the repository

abstract property version#

The repository version

Returns:

str

Return type:

the version of the repository

abstract property download_url#

The url template for downloading bundles

Returns:

str

Return type:

the download url

abstract property wildboar_requires#

The minimum required wildboar version

Returns:

str

Return type:

the min version

abstract get_bundles()#

Get all bundles

Returns:

dict

Return type:

a dictionary of key and bundle

get_bundle(key)#

Get a bundle with the specified key

Parameters:

key (str) – Key of the bundle

Returns:

bundle – A bundle or None

Return type:

Bundle, optional

load_dataset(bundle, dataset, *, cache_dir, version=None, tag=None, create_cache_dir=True, progress=True, dtype=None, force=False)#
list_datasets(bundle, *, cache_dir, version=None, tag=None, create_cache_dir=True, progress=True, force=False)#
clear_cache(cache_dir, keep_last_version=True)#
refresh()#

Refresh the repository

class wildboar.datasets.JSONRepository(url)#

Bases: Repository

A repository is a collection of bundles

property wildboar_requires#

The minimum required wildboar version

Returns:

str

Return type:

the min version

property name#

Name of the repository

Returns:

str

Return type:

the name of the repository

property version#

The repository version

Returns:

str

Return type:

the version of the repository

property download_url#

The url template for downloading bundles

Returns:

str

Return type:

the download url

get_bundles()#

Get all bundles

Returns:

dict

Return type:

a dictionary of key and bundle

refresh()#

Refresh the repository

wildboar.datasets.set_cache_dir(cache_dir)#

Change the global cache directory

cache_dirstr

The cache directory root

wildboar.datasets.load_synthetic_control(merge_train_test=True)#

Load the Synthetic_Control dataset

See also

load_dataset

load a named dataset

wildboar.datasets.load_two_lead_ecg(merge_train_test=True)#

Load the TwoLeadECG dataset

See also

load_dataset

load a named dataset

wildboar.datasets.load_gun_point(merge_train_test=True)#

Load the GunPoint dataset

See also

load_dataset

load a named dataset

wildboar.datasets.load_datasets(repository='wildboar/ucr', *, cache_dir=None, create_cache_dir=True, progress=True, force=False, filter=None, **kwargs)#

Load all datasets as a generator

Parameters:

  • repository (str) – The repository string

  • progress (bool, optional) – If progress indicator is shown while downloading the repository.

  • cache_dir (str, optional) – The cache directory for downloaded dataset repositories.

  • create_cache_dir (bool, optional) – Create the cache directory if it does not exist.

  • force (bool, optional) – Force re-download of cached repository

  • filter (dict or callable, optional) –

    Filter the datasets

    • if callable, only yield those datasets for which the callable returns True. f(dataset, x, y) -> bool

    • if dict, filter based on the keys and values

      • ”dataset”: regex matching dataset name

      • ”n_samples”: comparison spec

      • ”n_timestep”: comparison spec

    Comparison spec

    str of two parts, comparison operator (<, <=, >, >= or =) and a number, e.g., “<100”, “<= 200”, or “>300”

  • kwargs (dict) – Optional arguments to load_dataset

Yields:

  • x (array-like) – Data samples

  • y (array-like) – Data labels

Examples

>>> from wildboar.datasets import load_datasets
>>> for dataset, (x, y) in load_datasets(repository='wildboar/ucr'):
>>>     print(dataset, x.shape, y.shape)

Print the names of datasets with more than 200 samples

>>> for dataset, (x, y) in load_datasets(repository='wildboar/ucr', filter={"n_samples": ">200"}):
>>>     print(dataset)
wildboar.datasets.load_all_datasets(repository='wildboar/ucr', *, cache_dir=None, create_cache_dir=True, progress=True, force=False, **kwargs)#
wildboar.datasets.load_dataset(name, *, repository='wildboar/ucr', dtype=None, contiguous=True, merge_train_test=True, cache_dir=None, create_cache_dir=True, progress=True, force=False)#

Load a dataset from a repository

Parameters:

  • name (str) – The name of the dataset to load.

  • repository (str, optional) – The data repository formatted as {repository}/{bundle}[:{version}][:{tag}]

  • dtype (dtype, optional) – The data type of the returned data

  • contiguous (bool, optional) – Ensure that the returned dataset is memory contiguous.

  • merge_train_test (bool, optional) – Merge the existing training and testing partitions.

  • progress (bool, optional) – Show a progress bar while downloading a bundle.

  • cache_dir (str, optional) – The directory where downloaded files are cached

  • create_cache_dir (bool, optional) – Create cache directory if missing (default=True)

  • force (bool, optional) –

    Force re-download of already cached bundle

    ..versionadded :: 1.0.4

Returns:

  • x (ndarray) – The data samples

  • y (ndarray) – The labels

  • x_train (ndarray, optional) – The training samples if merge_train_test=False

  • x_test (ndarray, optional) – The testing samples if merge_train_test=False

  • y_train (ndarray, optional) – The training labels if merge_train_test=False

  • y_test (ndarray, optional) – The testing labels if merge_train_test=False

Examples

Load a dataset from the default repository

>>> x, y = load_dataset("SyntheticControl")

or if original training and testing splits are to be preserved

>>> x_train, x_test, y_train, y_test = load_dataset("SyntheticControl", merge_train_test=False)

or for a specific version of the dataset

>>> x_train, x_test, y_train, y_test = load_dataset("Wafer", repository='wildboar/ucr-tiny:1.0')
wildboar.datasets.list_datasets(repository='wildboar/ucr', *, cache_dir=None, create_cache_dir=True, progress=True, force=False)#

List the datasets in the repository

Parameters:

  • repository (str or Bundle, optional) –

    The data repository

    • if str load a named bundle, format {repository}/{bundle}

  • progress (bool, optional) – Show a progress bar while downloading a bundle.

  • cache_dir (str, optional) – The directory where downloaded files are cached (default=’wildboar_cache’)

  • create_cache_dir (bool, optional) – Create cache directory if missing (default=True)

  • force (bool, optional) – Force re-download of cached bundle

Returns:

dataset – A set of dataset names

Return type:

set

wildboar.datasets.clear_cache(repository=None, *, cache_dir=None, keep_last_version=True)#

Clear the cache by deleting cached datasets

Parameters:

  • repository (str, optional) –

    The name of the repository to clear cache.

    • if None, clear cache of all repositories

  • cache_dir (str, optional) – The cache directory

  • keep_last_version (bool, optional) – If true, keep the latest version of each repository.

wildboar.datasets.get_repository(repository)#

Get repository by name

Parameters:

repository (str) – Repository name

Returns:

repository – A repository

Return type:

Repository

wildboar.datasets.install_repository(repository)#

Install repository

Parameters:

repository (str or Repository) – A repository

wildboar.datasets.get_bundles(repository)#

Get all bundles in the repository

Parameters:

repository (str) – Name of the repository

Returns:

dict

Return type:

A dict of key Bundle pairs

wildboar.datasets.list_bundles(repository)#

Get a list of all bundle names in the specified repository.

Parameters:

repository (str) – The name of the repository

Returns:

bundle – The name of the bundle

Return type:

str

wildboar.datasets.list_repositories()#

List the key of all installed repositories