pyriemann.transfer.TLRotate¶
- class pyriemann.transfer.TLRotate(target_domain, weights=None, metric='euclid', n_jobs=1)¶
Rotate data for transfer learning.
Rotate the data points from each source domain so to match its class means with those from the target domain. The loss function for this matching was first proposed in [1] and the optimization procedure for mininimizing it follows the presentation from [2].
Note
The data points from each domain must have been re-centered to the identity before calculating the rotation.
Note
Using .fit() and then .transform() will give different results than .fit_transform(). In fact, .fit_transform() should be applied on the training dataset (target and source) and .transform() on the test partition of the target dataset.
- Parameters
- target_domainstr
Domain to consider as target.
- weightsNone | array, shape (n_classes,), default=None
Weights to assign for each class. If None, then give the same weight for each class.
- metric{‘euclid’, ‘riemann’}, default=’euclid’
Metric for the distance to minimize between class means. Options are either the Euclidean (‘euclid’) or Riemannian (‘riemann’) distance.
- n_jobsint, default=1
The number of jobs to use for the computation. This works by computing the rotation matrix for each source domain in parallel. If -1 all CPUs are used.
Notes
New in version 0.3.1.
References
- 1
Riemannian Procrustes analysis: transfer learning for brain-computer interfaces PLC Rodrigues et al, IEEE Transactions on Biomedical Engineering, vol. 66, no. 8, pp. 2390-2401, December, 2018
- 2
An introduction to optimization on smooth manifolds N. Boumal. To appear with Cambridge University Press. June, 2022
- Attributes
- rotations_dict
Dictionary with key=domain_name and value=domain_rotation_matrix.
- __init__(target_domain, weights=None, metric='euclid', n_jobs=1)¶
Init
- fit(X, y_enc)¶
Fit TLRotate.
Calculate the rotations matrices to transform each source domain into the target domain.
- Parameters
- Xndarray, shape (n_matrices, n_channels, n_channels)
Set of SPD matrices.
- y_encndarray, shape (n_matrices,)
Extended labels for each matrix.
- Returns
- selfTLRotate instance
The TLRotate instance.
- fit_transform(X, y_enc)¶
Fit TLRotate and then transform data points.
Calculate the rotation matrix for matching each source domain to the target domain.
Note
This method is designed for using at training time. The output for .fit_transform() will be different than using .fit() and .transform() separately.
- Parameters
- Xndarray, shape (n_matrices, n_channels, n_channels)
Set of SPD matrices.
- y_encndarray, shape (n_matrices,)
Extended labels for each matrix.
- Returns
- Xndarray, shape (n_matrices, n_classes)
Set of SPD matrices after rotation step.
- get_params(deep=True)¶
Get parameters for this estimator.
- Parameters
- deepbool, default=True
If True, will return the parameters for this estimator and contained subobjects that are estimators.
- Returns
- paramsdict
Parameter names mapped to their values.
- set_params(**params)¶
Set the parameters of this estimator.
The method works on simple estimators as well as on nested objects (such as
Pipeline
). The latter have parameters of the form<component>__<parameter>
so that it’s possible to update each component of a nested object.- Parameters
- **paramsdict
Estimator parameters.
- Returns
- selfestimator instance
Estimator instance.
- transform(X, y_enc=None)¶
Rotate the data points in the target domain.
The rotations are done from source to target, so in this step the data points suffer no transformation at all.
- Parameters
- Xndarray, shape (n_matrices, n_channels, n_channels)
Set of SPD matrices.
- y_encNone
Not used, here for compatibility with sklearn API.
- Returns
- Xndarray, shape (n_matrices, n_classes)
Same set of SPD matrices as in the input.