skada.deep.losses.dan_loss

skada.deep.losses.dan_loss(features_s, features_t, sigmas=None, eps=1e-07)

Define the mmd loss based on multi-kernel defined in [14].

Parameters:

features_stensor

Source features used to compute the mmd loss.

features_ttensor

Target features used to compute the mmd loss.

sigmasarray like, default=None,

If array, sigmas used for the multi gaussian kernel. If None, uses sigmas proposed in [1]_.

epsfloat, default=1e-7

Small constant added to median distance calculation for numerical stability.

Returns:

lossfloat

The loss of the method.

References

[14]

Mingsheng Long et. al. Learning Transferable Features with Deep Adaptation Networks. In ICML, 2015.