adapt.metrics.linear_discrepancy

adapt.metrics.linear_discrepancy(Xs, Xt, power_method=False, n_iter=20)

Compute the linear discrepancy between Xs and Xt.

\[\Delta = \max_{u \in \mathbb{R}^p} u^T (X_S^T X_S - X_T^T X_T) u\]

Where:

• \(p\) is the number of features of Xs and Xt.

Parameters

Xsarray

Source array

Xtarray

Target array

power_methodbool (default=False)

Weither to use the power method approximation or not.

n_iterint (default=20)

Number of iteration for power method

Returns

scorefloat

See also

normalized_linear_discrepancy

frechet_distance

normalized_frechet_distance

References

1

[1] Y. Mansour, M. Mohri, and A. Rostamizadeh. “Domain adaptation: Learning bounds and algorithms”. In COLT, 2009.