M/EEG signal decomposition using the Common Spatial Patterns (CSP). This class can be used as a supervised decomposition to estimate spatial filters for feature extraction. CSP in the context of EEG was first described in :footcite:`KolesEtAl1990`; a comprehensive tutorial on CSP can
| 24 | |
| 25 | @fill_doc |
| 26 | class CSP(_GEDTransformer): |
| 27 | """M/EEG signal decomposition using the Common Spatial Patterns (CSP). |
| 28 | |
| 29 | This class can be used as a supervised decomposition to estimate spatial |
| 30 | filters for feature extraction. CSP in the context of EEG was first |
| 31 | described in :footcite:`KolesEtAl1990`; a comprehensive tutorial on CSP can |
| 32 | be found in :footcite:`BlankertzEtAl2008`. Multi-class solving is |
| 33 | implemented from :footcite:`Grosse-WentrupBuss2008`. |
| 34 | |
| 35 | Parameters |
| 36 | ---------- |
| 37 | n_components : int (default 4) |
| 38 | The number of components to decompose M/EEG signals. This number should |
| 39 | be set by cross-validation. |
| 40 | reg : float | str | None (default None) |
| 41 | If not None (same as ``'empirical'``, default), allow regularization |
| 42 | for covariance estimation. If float (between 0 and 1), shrinkage is |
| 43 | used. For str values, ``reg`` will be passed as ``method`` to |
| 44 | :func:`mne.compute_covariance`. |
| 45 | log : None | bool (default None) |
| 46 | If ``transform_into`` equals ``'average_power'`` and ``log`` is None or |
| 47 | True, then apply a log transform to standardize features, else features |
| 48 | are z-scored. If ``transform_into`` is ``'csp_space'``, ``log`` must be |
| 49 | None. |
| 50 | cov_est : 'concat' | 'epoch' (default 'concat') |
| 51 | If ``'concat'``, covariance matrices are estimated on concatenated |
| 52 | epochs for each class. If ``'epoch'``, covariance matrices are |
| 53 | estimated on each epoch separately and then averaged over each class. |
| 54 | transform_into : 'average_power' | 'csp_space' (default 'average_power') |
| 55 | If 'average_power' then ``self.transform`` will return the average |
| 56 | power of each spatial filter. If ``'csp_space'``, ``self.transform`` |
| 57 | will return the data in CSP space. |
| 58 | norm_trace : bool (default False) |
| 59 | Normalize class covariance by its trace. Trace normalization is a step |
| 60 | of the original CSP algorithm :footcite:`KolesEtAl1990` to eliminate |
| 61 | magnitude variations in the EEG between individuals. It is not applied |
| 62 | in more recent work :footcite:`BlankertzEtAl2008`, |
| 63 | :footcite:`Grosse-WentrupBuss2008` and can have a negative impact on |
| 64 | pattern order. |
| 65 | cov_method_params : dict | None |
| 66 | Parameters to pass to :func:`mne.compute_covariance`. |
| 67 | |
| 68 | .. versionadded:: 0.16 |
| 69 | |
| 70 | restr_type : "restricting" | "whitening" | None |
| 71 | Restricting transformation for covariance matrices before performing |
| 72 | generalized eigendecomposition. |
| 73 | If "restricting" only restriction to the principal subspace of signal_cov |
| 74 | will be performed. |
| 75 | If "whitening", covariance matrices will be additionally rescaled according |
| 76 | to the whitening for the signal_cov. |
| 77 | If None, no restriction will be applied. Defaults to "restricting". |
| 78 | |
| 79 | .. versionadded:: 1.11 |
| 80 | info : mne.Info | None |
| 81 | The mne.Info object with information about the sensors and methods of |
| 82 | measurement used for covariance estimation and generalized |
| 83 | eigendecomposition. |
no outgoing calls