class WhiteningTransform

class deeptime.covariance.WhiteningTransform(sqrt_inv_cov: ndarray, mean: Optional[ndarray] = None, dim: Optional[int] = None)

Transformation of data into a whitened space. It is assumed that for a covariance matrix \(C\) the square-root inverse \(C^{-1/2}\) was already computed. Optionally a mean \(\mu\) can be provided. This yields the transformation

\[y = C^{-1/2}(x-\mu). \]

Parameters:

• sqrt_inv_cov ((n, k) ndarray) – Square-root inverse of covariance matrix.

• mean ((n, ) ndarray, optional, default=None) – The mean if it should be subtracted.

• dim (int, optional, default=None) – Additional restriction in the dimension, removes all but the first dim components of the output.

See also

deeptime.numeric.spd_inv_sqrt

Method to obtain (regularized) inverses of covariance matrices.

Methods

transform(data, **kwargs)

Transforms the input data.

__call__(x: ndarray)

Evaluation of the observable.

Parameters:

x ((N, d) np.ndarray) – Evaluates the observable for N d-dimensional data points.

Returns:

out – Result of the evaluation for each data point.

Return type:

(N, p) np.ndarray

transform(data, **kwargs)

Transforms the input data.

Parameters:

data (array_like) – Input data.

Returns:

transformed – The transformed data

Return type:

array_like