holisticai.bias.mitigation.PopularityPropensityMF

class holisticai.bias.mitigation.PopularityPropensityMF(*args, **kargs)

Popularity Propensity Matrix Factorization [1] address selection biases in recommender systems by using causal inference techniques to provide unbiased performance estimators and improve prediction accuracy. This method estimates the probability (propensity) that a user will rate an item and adjusts the training and evaluation processes accordingly.

Parameters

Kint

Specifies the number of dimensions.

betafloat

Parameter used to update P and Q.

stepsint

Number of iterations.

verboseint

If >0, will show progress percentage.

Examples

>>> from holisticai.bias.mitigation import PopularityPropensityMF
>>> mitigator = PopularityPropensityMF(**params)
>>> mitigator.fit(data_matrix)

References

[1]

Tobias Schnabel, Adith Swaminathan, Ashudeep Singh, Navin Chandak, and Thorsten Joachims. 2016. Recommendations as treatments: Debiasing learning and evaluation. arXiv preprint arXiv:1602.05352 (2016).

fit(X: ndarray | None, **kargs)

Fit model

Parameters

Xmatrix-like

rating matrix, 0 means non-raked cases.

P0matrix-like (optional)

Initial P matrix (numUsers, K)

Q0matrix-like (optional)

Initial P matrix (numItems, K)

Returns

self