d3m.metadata.problem

class d3m.metadata.problem.PerformanceMetric(value)

Bases: d3m.metadata.problem.PerformanceMetricBase, d3m.utils.Enum

An enumeration.

ACCURACY = 'ACCURACY'

F1 = 'F1'

F1_MACRO = 'F1_MACRO'

F1_MICRO = 'F1_MICRO'

HAMMING_LOSS = 'HAMMING_LOSS'

HITS_AT_K = 'HITS_AT_K'

JACCARD_SIMILARITY_SCORE = 'JACCARD_SIMILARITY_SCORE'

MEAN_ABSOLUTE_ERROR = 'MEAN_ABSOLUTE_ERROR'

MEAN_RECIPROCAL_RANK = 'MEAN_RECIPROCAL_RANK'

MEAN_SQUARED_ERROR = 'MEAN_SQUARED_ERROR'

NORMALIZED_MUTUAL_INFORMATION = 'NORMALIZED_MUTUAL_INFORMATION'

OBJECT_DETECTION_AVERAGE_PRECISION = 'OBJECT_DETECTION_AVERAGE_PRECISION'

PRECISION = 'PRECISION'

PRECISION_AT_TOP_K = 'PRECISION_AT_TOP_K'

RECALL = 'RECALL'

ROC_AUC = 'ROC_AUC'

ROC_AUC_MACRO = 'ROC_AUC_MACRO'

ROC_AUC_MICRO = 'ROC_AUC_MICRO'

ROOT_MEAN_SQUARED_ERROR = 'ROOT_MEAN_SQUARED_ERROR'

R_SQUARED = 'R_SQUARED'

class d3m.metadata.problem.Problem(problem_description=None, *, strict_digest=False)

Bases: dict

A class representing a problem.

copy()

Return type

~P

classmethod from_json_structure(structure, *, strict_digest=False)

For standard enumerations we map them from strings. For non-standard problem description fields we used a reversible conversion.

Return type

~P

classmethod from_simple_structure(structure, *, strict_digest=False)

Return type

~P

get_digest()

Return type

str

classmethod load(problem_uri, *, problem_id=None, problem_version=None, problem_name=None, strict_digest=False, handle_score_split=True)

Tries to load problem from problem_uri using all registered problem loaders.

Parameters

• problem_uri (str) – A URI to load.

• problem_id (Optional[str]) – Override problem ID determined by the loader.

• problem_version (Optional[str]) – Override problem version determined by the loader.

• problem_name (Optional[str]) – Override problem name determined by the loader.

• strict_digest (bool) – If computed digest does not match the one provided in metadata, raise an exception?

• handle_score_split (bool) – Rename a scoring problem to not have the same name as testing problem and update dataset references.

Returns

A loaded problem.

Return type

Problem

classmethod register_loader(loader)

Registers a new problem loader.

Parameters

loader (Loader) – An instance of the loader class implementing a new loader.

Return type

None

classmethod register_saver(saver)

Registers a new dataset saver.

Parameters

saver (Saver) – An instance of the saver class implementing a new saver.

Return type

None

save(problem_uri)

Tries to save dataset to problem_uri using all registered problem savers.

Parameters

problem_uri (str) – A URI to save to.

Return type

None

to_json_structure(*, canonical=False)

For standard enumerations we map them to strings. Non-standard problem description fields we convert in a reversible manner.

Return type

Dict

to_simple_structure(*, canonical=False)

Return type

Dict

loaders: List[d3m.metadata.problem.Loader] = [<d3m.metadata.problem.D3MProblemLoader object>, <d3m.metadata.problem.OpenMLProblemLoader object>]

savers: List[d3m.metadata.problem.Saver] = [<d3m.metadata.problem.D3MProblemSaver object>]

class d3m.metadata.problem.TaskKeyword(value)

Bases: d3m.metadata.problem.TaskKeywordBase, d3m.utils.Enum

An enumeration.

AUDIO = 'AUDIO'

BINARY = 'BINARY'

CLASSIFICATION = 'CLASSIFICATION'

CLUSTERING = 'CLUSTERING'

COLLABORATIVE_FILTERING = 'COLLABORATIVE_FILTERING'

COMMUNITY_DETECTION = 'COMMUNITY_DETECTION'

FORECASTING = 'FORECASTING'

GEOSPATIAL = 'GEOSPATIAL'

GRAPH = 'GRAPH'

GRAPH_MATCHING = 'GRAPH_MATCHING'

GROUPED = 'GROUPED'

IMAGE = 'IMAGE'

LINK_PREDICTION = 'LINK_PREDICTION'

LUPI = 'LUPI'

MISSING_METADATA = 'MISSING_METADATA'

MULTICLASS = 'MULTICLASS'

MULTIGRAPH = 'MULTIGRAPH'

MULTILABEL = 'MULTILABEL'

MULTIPLE_INSTANCE_LEARNING = 'MULTIPLE_INSTANCE_LEARNING'

MULTIVARIATE = 'MULTIVARIATE'

NESTED = 'NESTED'

NONOVERLAPPING = 'NONOVERLAPPING'

OBJECT_DETECTION = 'OBJECT_DETECTION'

OVERLAPPING = 'OVERLAPPING'

REGRESSION = 'REGRESSION'

RELATIONAL = 'RELATIONAL'

REMOTE_SENSING = 'REMOTE_SENSING'

SEMISUPERVISED = 'SEMISUPERVISED'

SPEECH = 'SPEECH'

TABULAR = 'TABULAR'

TEXT = 'TEXT'

TIME_SERIES = 'TIME_SERIES'

UNIVARIATE = 'UNIVARIATE'

VERTEX_CLASSIFICATION = 'VERTEX_CLASSIFICATION'

VERTEX_NOMINATION = 'VERTEX_NOMINATION'

VIDEO = 'VIDEO'