quantus.helpers.enums module
- class quantus.helpers.enums.DataType(value)
Bases:
Enum
This enum represents the different types of data that a metric implementation currently supports.
IMAGE: Represents image data.
TABULAR: Represents tabular data.
TEXT: Represents text data.
- IMAGE = 'image'
- TABULAR = 'tabular'
- TEXT = 'text'
- TIMESERIES = 'time-series'
- class quantus.helpers.enums.EvaluationCategory(value)
Bases:
Enum
This enum represents different categories of explanation quality for XAI algorithms.
FAITHFULNESS: Indicates how well the explanation reflects the true features used by the model.
ROBUSTNESS: Represents the degree to which the explanation remains consistent under small perturbations in the input.
RANDOMISATION: Measures the quality of the explanation in terms of difference in explanation when randomness is introduced.
COMPLEXITY: Refers to how easy it is to understand the explanation. Lower complexity is usually better.
LOCALISATION: Refers to how consistently the explanation points out the parts of the input as defined in a ground-truth segmentation mask.
AXIOMATIC: Represents the quality of the explanation in terms of well-defined axioms.
- AXIOMATIC = 'Axiomatic'
- COMPLEXITY = 'Complexity'
- FAITHFULNESS = 'Faithfulness'
- LOCALISATION = 'Localisation'
- NONE = 'None'
- RANDOMISATION = 'Randomisation'
- ROBUSTNESS = 'Robustness'