Concept

class hybrid_learning.concepts.concepts.Concept(name, data)

Bases: object

Representation of a concept with data and meta information. Sub-classes should implement the property type and extend the data checks in _validate_dataset(). Different general types can be found in ConceptTypes.

Public Data Attributes:

type	The type of the concept including according type checks.
train_data	Training data set.
val_data	Validation data set.
test_data	Test data set.
train_val_data	Combined dataset of the defining training and validation data.
settings	Settings dict to reproduce instance.

Special Methods:

__init__(name, data)	Init.
__eq__(other)	Equality check by checking on values.
__repr__()	Nice printing function.

---

Parameters

• name (str) –

• data (DataTriple) –

__eq__(other)

Equality check by checking on values. Especially, datasets must be the same objects.

__init__(name, data)

Init.

Parameters

• name (str) – name and ID of the concept (should be descriptive and unique)

• data (DataTriple) – dataset splits of data representing the concept

__repr__()

Nice printing function.

Return type

str

static _validate_dataset(data, data_desc='??')

Check whether the given data fulfills all required properties and raise if not. Extend in sub-class.

Parameters

• data (torch.utils.data.Dataset) – dataset to validate

• data_desc (str) – some description of the data for more informative error messages

Raises

ValueError

Returns

the unchanged given dataset

Return type

torch.utils.data.Dataset

classmethod new(concept)

Initialize a concept from the attributes of another by references. This is helpful when converting concepts to specific sub-classes (with the corresponding data checks during init).

Parameters

concept (Concept) –

__hash__ = None

data: DataTriple

The data defining the concept, split into train/val/test data.

name: str

The name, or ID, of the concept. Should be descriptive.

property settings: Dict[str, Any]

Settings dict to reproduce instance. Use as keyword args for init.

property test_data: torch.utils.data.Dataset

Test data set. See test attribute of data.

property train_data: torch.utils.data.Dataset

Training data set. See train attribute of data.

property train_val_data: torch.utils.data.Dataset

Combined dataset of the defining training and validation data. See train_val attribute of data.

property type: ConceptTypes

The type of the concept including according type checks.

property val_data: torch.utils.data.Dataset

Validation data set. See val attribute of data.