bqlearn.metrics.iterative_anchor_transition_matrix

bqlearn.metrics.iterative_anchor_transition_matrix(y_prob, quantile=0.97, n_iter=100)

Compute a transition matrix based on an iterative algorithm using anchor points [1].

Parameters:

y_probarray-like of shape (n_samples, n_classes)

Predicted probabilities, as returned by a classifier’s predict_proba method.

quantilefloat, default=0.97

Quantile used to select the anchor points. It filters out outlier points with high predicted probabilities.

n_iterint, default=100

Number of time an enhanced anchor transition matrix is computed.

Returns:

Cndarray of shape (n_classes, n_classes)

Anchor transition matrix whose i-th row and j-th column entry indicates the probability of samples with true label being i-th class to be corrupted to a label being the j-th class.

References

[1]

M. Zhang, J. Lee, and S. Agarwal. “Learning from noisy labels with no change to the training process.”, ICML, 2021.