lib.sedna.algorithms.unseen_task_detect.unseen_task_detect

Unseen task detection algorithms for Lifelong Learning

Module Contents

Classes

ModelProbeFilter	Judgment based on the confidence of the prediction result,
TaskAttrFilter	Judgment based on whether the metadata of the sample has been found in KB

class lib.sedna.algorithms.unseen_task_detect.unseen_task_detect.ModelProbeFilter

Bases: BaseFilter, abc.ABC

Judgment based on the confidence of the prediction result, typically used for classification problems

__call__(tasks: List[sedna.algorithms.multi_task_learning.task_jobs.artifact.Task] = None, threshold=0.5, **kwargs)

Parameters:

• tasks (inference task) –

• threshold (float) – threshold considered credible

Returns:

is unseen task – True means unseen task, False means not.

Return type:

bool

class lib.sedna.algorithms.unseen_task_detect.unseen_task_detect.TaskAttrFilter

Bases: BaseFilter, abc.ABC

Judgment based on whether the metadata of the sample has been found in KB

__call__(tasks: List[sedna.algorithms.multi_task_learning.task_jobs.artifact.Task] = None, **kwargs)

Parameters:

tasks (inference task) –

Returns:

is unseen task – True means unseen task, False means not.

Return type:

bool