lenskit.reranking.fair.FAIRRerankerConfig

class lenskit.reranking.fair.FAIRRerankerConfig(/, **data)

Bases: pydantic.BaseModel

Configuration for FAIRReranker.

Parameters:

data (Any)

n: pydantic.PositiveInt

Size of the top-N subset to rerank. If None, the entire candidate list is reranked.

p: pydantic.PositiveFloat

The target proportion of protected items in the list (0.0 ≤ p ≤ 1.0). The default value is 0.5.

alpha: pydantic.PositiveFloat

The probability threshold for Type I error across prefixes. The default value is 0.1.

protected_attribute: str

The item attribute name that marks whether an item is protected. The default name is ‘protected’.