himalaya.scoring.correlation_score_split

himalaya.scoring.correlation_score_split(y_true, y_pred)[source]

Split the correlation score into individual components.

When estimating a linear joint model, the predictions of each feature space are summed:

Yhat_joint = Yhat_A + Yhat_B + ... + Yhat_Z

The joint model correlation score r can be computed as:

r_joint = r(Y, Yhat_joint)

This function estimates the contribution of each feature space to the joint model correlation score r such that:

r_joint = r_A + r_B + ... + r_Z
Parameters
y_truearray or Tensor of shape (n_samples, n_targets)

Ground truth.

y_predarray or Tensor of shape (n_predictions, n_samples, n_targets) or (n_samples, n_targets)

Predictions.

Returns
correlationsarray of shape (n_predictions, n_targets) or (n_targets, )

Contributions of each individual feature space to the joint correlation score.