skada.deep.MCCLoss

class skada.deep.MCCLoss(T=1, eps=1e-07)

Loss MCC.

This loss reduces the class confusion of the predicted label of target domain See [33].

Parameters:

Tfloat, default=1

Temperature parameter for the scaling. If T=1, the scaling is a softmax function.

epsfloat, default=1e-7

Small constant added to median distance calculation for numerical stability.

References

[33]

Ying Jin, Ximei Wang, Mingsheng Long, Jianmin Wang. Minimum Class Confusion for Versatile Domain Adaptation. In ECCV, 2020.

forward(y_pred_t, **kwargs)

Compute the domain adaptation loss