Aluode/PerceptionLabPortable
0
1o
2 '�Yi { � @ sZ d dl Zd dlmZmZmZmZ d dlmZm Z d dl3m Z d dl m
Z
ddlmZmZmZmZ ddlmZ dd lmZ dd4lmZ dd lmZ dd lm Z m Z m Z m Z d
dl m!Z! G dd� de�Z"G dd� d�Z#dd� Z$e G dd� de"e��Z%G dd� de"e�Z&e G dd� de"e��Z'e G dd � d e"e��Z(G d d � d e"e�Z)e G d d � d e"e��Z*dS )!� N)�
BaseEstimator�TransformerMixin� check_array�clone)� RobustScaler�StandardScaler)� check_X_y)�check_is_fitted� )�_pick_data_channels�_picks_by_type�
_picks_to_idx� pick_info)�_check_scalings_user)�5BaseEpochs)� filter_data)�psd_array_multitaper)�
_check_option�_validate_type�
check_version�fill_doc� )�
validate_datac @ s( e Zd ZdZddddddd�dd�ZdS )�MNETransformerMixinz#TransformerMixin plus some helpers.NTF)�y�6atleast_3d�fit�return_y� multi_output�check_n_featuresc C s� t |t�r |jdd�}ttjddd�}tdd�r d|d< t| d �rF|rF|d u r5t| |fi |�d7|i��}n+t| ||fi |�d8|i��\}}n|d u rSt |fi |��}n
t9d
|||d �|��\}}|rh|j d | _ |rot�
|�}|ru||fS |S )NF)�copyT�C)ZdtypeZallow_nd�orderZsklearnz1.4Zforce_writeable�n_features_in_�reset)�Xr r r � )�10isinstancer �get_data�dict�npZfloat64r �hasattrr r r �shaper# r ) �self� epochs_datar r r r r r �kwargsr& r& ��E:\DocsHouse\542 percep lab latest\PerceptionLab\PerceptionLab_Portable\python_embed\Lib\site-packages\mne/decoding/transformer.py� _check_data sH 11 12��13�����14� 15z MNETransformerMixin._check_data)�__name__�16__module__� __qualname__�__doc__r1 r& r& r& r0 r s �r c @ s@ e Zd ZdZddd�Zddd�Zdd � Zdd17d �Zdd d
�ZdS )�_ConstantScalerz*Scale channel types using constant values.Tc C s || _ || _|| _d S �N)� _scalings�_info� _do_scaling)r- �info�scalingsZ18do_scalingr& r& r0 �__init__P s 19z_ConstantScaler.__init__Nc C s� t | j�}tt| jt| jdd���}t�tdd� |D ���}|j d t20|�kr6t dt21|�� dt22|�� d���| j rH|D ] \}}d || ||< q;|| _
t�|�| _| S )23Nr& )�excludec s s � | ] }t |d �V qdS )r N)�len)�.0�pr& r& r0 � <genexpr>Z s � z&_ConstantScaler.fit.<locals>.<genexpr>r z info had z data channels but X has z channels� �?)r r8 r r r9 r r* Zones�sumr, r? �24ValueErrorr: �std_Z25zeros_likeZmean_)r- r% r r< Z
picks_by_typeZstd�kind�picksr& r& r0 r U s 26�� z_ConstantScaler.fitc C s27 || j S r7 �rF �r- r% r& r& r0 � transformf � 28z_ConstantScaler.transformc C s29 || j S r7 rI �r- r% r r& r& r0 �inverse_transformi rL z!_ConstantScaler.inverse_transformc C � | � ||��|�S r7 �r rK rM r& r& r0 �
fit_transforml s z _ConstantScaler.fit_transform)Tr7 ) r2 r3 r4 r5 r= r rK rN rQ r& r& r& r0 r6 M s 303132r6 c O s� t |tjd� |jdkr|r|�� S dS |j}t�|�ddd�d|d f�}| |g|�R i |��}|rH|d |d |d f|_|�ddd�}|S dS )z"Reshape epochs and apply function.r% r Nr33 r �����)r r* Zndarray�sizer r, �reshape� transpose)�funcZ
return_resultr% �argsr/ Z34orig_shaper& r& r0 �_sklearn_reshape_applyp s 35 �rX c @ s> e Zd ZdZddd�Zddd�Zdd � Zdd36d �Zd d
� ZdS )�Scaleru� Standardize channel data.37 38 This class scales data for each channel. It differs from scikit-learn39 classes (e.g., :class:`sklearn.preprocessing.StandardScaler`) in that40 it scales each *channel* by estimating μ and σ using data from all41 time points and epochs, as opposed to standardizing each *feature*42 (i.e., each time point for each channel) by estimating using μ and σ43 using data from all epochs.44 45 Parameters46 ----------47 %(info)s Only necessary if ``scalings`` is a dict or None.48 scalings : dict, str, default None49 Scaling method to be applied to data channel wise.50 51 * if scalings is None (default), scales mag by 1e15, grad by 1e13,52 and eeg by 1e6.53 * if scalings is :class:`dict`, keys are channel types and values54 are scale factors.55 * if ``scalings=='median'``,56 :class:`sklearn.preprocessing.RobustScaler`57 is used (requires sklearn version 0.17+).58 * if ``scalings=='mean'``,59 :class:`sklearn.preprocessing.StandardScaler`60 is used.61 62 with_mean : bool, default True63 If True, center the data using mean (or median) before scaling.64 Ignored for channel-type scaling.65 with_std : bool, default True66 If True, scale the data to unit variance (``scalings='mean'``),67 quantile range (``scalings='median``), or using channel type68 if ``scalings`` is a dict or None).69 NTc C s || _ || _|| _|| _d S r7 )r; � with_mean�with_stdr< )r- r; r< rZ r[ r& r& r0 r= � s 70zScaler.__init__c C s� | j ||ddd�}|jdksJ |j��t| jtttd�fd� t| jt�r0t d| jddgdd � | jdu s;t| jt�rU| j71du rJt d72t| j�� ���t | j73| j| j
�| _n| jdkrdt| j| j
d �| _n t| j| j
d �| _t| jjd
||d� | S )as Standardize data across channels.74 75 Parameters76 ----------77 epochs_data : array, shape (n_epochs, n_channels, n_times)78 The data to concatenate channels.79 y : array, shape (n_epochs,)80 The label for each epoch.81 82 Returns83 -------84 self : instance of Scaler85 The modified instance.86 T)r r r � Nr< �meanZmedianzwhen str)�extraz&Need to specify "info" if scalings is )rZ r[ )Zwith_centeringZ with_scalingF)r )r1 �ndimr, r r< r) �str�typer' r r; rE r6 r[ �scaler_r rZ r rX r �r- r. r r& r& r0 r � s, �87�8889��z90Scaler.fitc C sz t | d� | j|dd�}|jdkr+| jdur$t| jd �|jd ks$J �|dtjf }|jd ks5J |j��t| j j91d92|�S ) a� Standardize data across channels.93 94 Parameters95 ----------96 epochs_data : array, shape (n_epochs, n_channels[, n_times])97 The data.98 99 Returns100 -------101 X : array, shape (n_epochs, n_channels, n_times)102 The data concatenated over channels.103 104 Notes105 -----106 This function makes a copy of the data before the operations and the107 memory usage may be large with big data.108 rb F�r r109 NZch_namesr .r\ T) r r1 r_ r; r? r, r* �newaxisrX rb rK �r- r. r& r& r0 rK � s 110111112 zScaler.transformc C rO )a� Fit to data, then transform it.113 114 Fits transformer to epochs_data and y and returns a transformed version115 of epochs_data.116 117 Parameters118 ----------119 epochs_data : array, shape (n_epochs, n_channels, n_times)120 The data.121 y : None | array, shape (n_epochs,)122 The label for each epoch.123 Defaults to None.124 125 Returns126 -------127 X : array, shape (n_epochs, n_channels, n_times)128 The data concatenated over channels.129 130 Notes131 -----132 This function makes a copy of the data before the operations and the133 memory usage may be large with big data.134 rP rc r& r& r0 rQ � s zScaler.fit_transformc C sb | j |dd�}d}|jdkrd}|dtjf }|jdks!J |j��t| jjd|�}|r/|d }|S )a� Invert standardization of data across channels.135 136 Parameters137 ----------138 epochs_data : array, shape ([n_epochs, ]n_channels, n_times)139 The data.140 141 Returns142 -------143 X : array, shape (n_epochs, n_channels, n_times)144 The data concatenated over channels.145 146 Notes147 -----148 This function makes a copy of the data before the operations and the149 memory usage may be large with big data.150 Frd r151 T.r\ ).r )r1 r_ r* re r, rX rb rN )r- r. Zsqueeze�outr& r& r0 rN s 152zScaler.inverse_transform)NNTTr7 ) r2 r3 r4 r5 r= r rK rQ rN r& r& r& r0 rY ~ s 153#154)155 rY c @ s4 e Zd ZdZd dd�Zdd� Zd dd�Zd d156� ZdS ) �157Vectorizera] Transform n-dimensional array into 2D array of n_samples by n_features.158 159 This class reshapes an n-dimensional array into an n_samples * n_features160 array, usable by the estimators and transformers of scikit-learn.161 162 Attributes163 ----------164 features_shape_ : tuple165 Stores the original shape of data.166 167 Examples168 --------169 >>> from sklearn.linear_model import LogisticRegression170 >>> from sklearn.pipeline import make_pipeline171 >>> from sklearn.preprocessing import StandardScaler172 >>> clf = make_pipeline(Vectorizer(), StandardScaler(), LogisticRegression())173 Nc C s( | j ||dddd�}|jdd� | _| S )aG Store the shape of the features of X.174 175 Parameters176 ----------177 X : array-like178 The data to fit. Can be, for example a list, or an array of at179 least 2d. The first dimension must be of length n_samples, where180 samples are the independent samples used by the estimator181 (e.g. n_epochs for epoched data).182 y : None | array, shape (n_samples,)183 Used for scikit-learn compatibility.184 185 Returns186 -------187 self : instance of Vectorizer188 Return the modified instance.189 FT)r r r r r N)r1 r, �features_shape_rM r& r& r0 r 9 s zVectorizer.fitc C s: | j |dd�}|jdd� | jkrtd��|�t|�d�S )a� Convert given array into two dimensions.190 191 Parameters192 ----------193 X : array-like194 The data to fit. Can be, for example a list, or an array of at195 least 2d. The first dimension must be of length n_samples, where196 samples are the independent samples used by the estimator197 (e.g. n_epochs for epoched data).198 199 Returns200 -------201 X : array, shape (n_samples, n_features)202 The transformed data.203 Frd r Nz1Shape of X used in fit and transform must be samerR )r1 r, ri rE rT r? rJ r& r& r0 rK O s zVectorizer.transformc C � | � |��|�S )aF Fit the data, then transform in one step.204 205 Parameters206 ----------207 X : array-like208 The data to fit. Can be, for example a list, or an array of at209 least 2d. The first dimension must be of length n_samples, where210 samples are the independent samples used by the estimator211 (e.g. n_epochs for epoched data).212 y : None | array, shape (n_samples,)213 Used for scikit-learn compatibility.214 215 Returns216 -------217 X : array, shape (n_samples, -1)218 The transformed data.219 rP rM r& r& r0 rQ d s zVectorizer.fit_transformc C sD | j |ddd�}|jdvrtd|j� ���|�|jdd� | j �S )a~ Transform 2D data back to its original feature shape.220 221 Parameters222 ----------223 X : array-like, shape (n_samples, n_features)224 Data to be transformed back to original shape.225 226 Returns227 -------228 X : array229 The data transformed into shape as used in fit. The first230 dimension is of length n_samples.231 F)r r )r232 r\ z/X should be of 2 or 3 dimensions but has shape NrR )r1 r_ rE r, rT ri rJ r& r& r0 rN x s 233234�z Vectorizer.inverse_transformr7 )r2 r3 r4 r5 r rK rQ rN r& r& r& r0 rh & s 235236 rh c sT e Zd ZdZdej dejdddddfdd �Z� fd237d �Zdd d
�Z dd� Z238� Z S )� PSDEstimatoraj Compute power spectral density (PSD) using a multi-taper method.239 240 Parameters241 ----------242 sfreq : float243 The sampling frequency.244 fmin : float245 The lower frequency of interest.246 fmax : float247 The upper frequency of interest.248 bandwidth : float249 The bandwidth of the multi taper windowing function in Hz.250 adaptive : bool251 Use adaptive weights to combine the tapered spectra into PSD252 (slow, use n_jobs >> 1 to speed up computation).253 low_bias : bool254 Only use tapers with more than 90%% spectral concentration within255 bandwidth.256 n_jobs : int257 Number of parallel jobs to use (only used if adaptive=True).258 %(normalization)s259 260 See Also261 --------262 mne.time_frequency.psd_array_multitaper263 mne.io.Raw.compute_psd264 mne.Epochs.compute_psd265 mne.Evoked.compute_psd266 r267 r NFT�lengthc C s4 || _ || _|| _|| _|| _|| _|| _|| _d S r7 )�sfreq�fmin�fmax� bandwidth�adaptive�low_bias�n_jobs�
normalization) r- rm rn ro rp rq rr rs rt r& r& r0 r= � s 268zPSDEstimator.__init__c � t � �� }d|j_d|_|S �z...F��super�__sklearn_tags__Z target_tags�requiredZ requires_fit�r- �tags�� __class__r& r0 ry � � 269z PSDEstimator.__sklearn_tags__c C s | j ||dd� d| _| S )a� Compute power spectral density (PSD) using a multi-taper method.270 271 Parameters272 ----------273 epochs_data : array, shape (n_epochs, n_channels, n_times)274 The data.275 y : array, shape (n_epochs,)276 The label for each epoch.277 278 Returns279 -------280 self : instance of PSDEstimator281 The modified instance.282 T�r r )r1 �fitted_rc r& r& r0 r � s zPSDEstimator.fitc C s< | � |�}t|| j| j| j| j| j| j| j| j d� \}}|S )aF Compute power spectral density (PSD) using a multi-taper method.283 284 Parameters285 ----------286 epochs_data : array, shape (n_epochs, n_channels, n_times)287 The data.288 289 Returns290 -------291 psd : array, shape (n_signals, n_freqs) or (n_freqs,)292 The computed PSD.293 )rm rn ro rp rq rr rt rs )294r1 r rm rn ro rp rq rr rt rs )r- r. Zpsd�_r& r& r0 rK � s 295
296� zPSDEstimator.transformr7 ) r2 r3 r4 r5 r* �pi�infr= ry r rK �
__classcell__r& r& r} r0 rk � s 297� 298rk c @ s: e Zd ZdZ d dd�Zdd � Zd299d � ZdS )
�FilterEstimatora Estimator to filter RtEpochs.300 301 Applies a zero-phase low-pass, high-pass, band-pass, or band-stop302 filter to the channels selected by "picks".303 304 l_freq and h_freq are the frequencies below which and above which,305 respectively, to filter out of the data. Thus the uses are:306 307 - l_freq < h_freq: band-pass filter308 - l_freq > h_freq: band-stop filter309 - l_freq is not None, h_freq is None: low-pass filter310 - l_freq is None, h_freq is not None: high-pass filter311 312 If n_jobs > 1, more memory is required as "len(picks) * n_times"313 additional time points need to be temporarily stored in memory.314 315 Parameters316 ----------317 %(info_not_none)s318 %(l_freq)s319 %(h_freq)s320 %(picks_good_data)s321 %(filter_length)s322 %(l_trans_bandwidth)s323 %(h_trans_bandwidth)s324 n_jobs : int | str325 Number of jobs to run in parallel.326 Can be 'cuda' if ``cupy`` is installed properly and method='fir'.327 method : str328 'fir' will use overlap-add FIR filtering, 'iir' will use IIR filtering.329 iir_params : dict | None330 Dictionary of parameters to use for IIR filtering.331 See mne.filter.construct_iir_filter for details. If iir_params332 is None and method="iir", 4th order Butterworth will be used.333 %(fir_design)s334 335 See Also336 --------337 TemporalFilter338 339 Notes340 -----341 This is primarily meant for use in realtime applications.342 In general it is not recommended in a normal processing pipeline as it may result343 in edge artifacts. Use with caution.344 N�auto�fir�firwinc C �F || _ || _|| _|| _|| _|| _|| _|| _| | _|345| _ | | _346d S r7 ) r; �l_freq�h_freqrH �
filter_length�l_trans_bandwidth�h_trans_bandwidthrs �method�347iir_params�348fir_design) r- r; r� r� rH r� r� r� rs r� r� r� r& r& r0 r= ) � 349zFilterEstimator.__init__c C s t | j| j�| _| j||dd� | jdkrd| _| jd du s7| jdurR| jdu s/| j| jk rR| j| jd k rR| j�� � | j| jd< W d � n1 sMw Y | jd du sq| jdur�| jdu si| j| jk r�| j| jd kr�| j�� � | j| jd< W d � | S 1 s�w Y | S )aO Filter data.350 351 Parameters352 ----------353 epochs_data : array, shape (n_epochs, n_channels, n_times)354 The data.355 y : array, shape (n_epochs,)356 The label for each epoch.357 358 Returns359 -------360 self : instance of FilterEstimator361 The modified instance.362 Tr� r NZlowpassZhighpass)r
r; rH �picks_r1 r� r� Z_unlockrc r& r& r0 r C s* 363364 �365 366��zFilterEstimator.fitc C sD t | �|�| jd | j| j| j| j| j| j| j | j367| j d| j dd�S )a Filter data.368 369 Parameters370 ----------371 epochs_data : array, shape (n_epochs, n_channels, n_times)372 The data.373 374 Returns375 -------376 X : array, shape (n_epochs, n_channels, n_times)377 The data after filtering.378 rm F)r� r� rs r r� �verbose)
r r1 r; r� r� r� r� r� r� r� r� rs r� rf r& r& r0 rK j s
�zFilterEstimator.transform)Nr� r� r� Nr� Nr� )r2 r3 r4 r5 r= r rK r& r& r& r0 r� � s 4379� 'r� c @ sF e Zd ZdZddd�Zddd�Zddd �Zd380d � Zd d
� Zdd� Z dS )�UnsupervisedSpatialFilteraL Use unsupervised spatial filtering across time and samples.381 382 Parameters383 ----------384 estimator : instance of sklearn.base.BaseEstimator385 Estimator using some decomposition algorithm.386 average : bool, default False387 If True, the estimator is fitted on the average across samples388 (e.g. epochs).389 Fc C s || _ || _d S r7 )� estimator�average)r- r� r� r& r& r0 r= � s 390z"UnsupervisedSpatialFilter.__init__Nc C s� dD ]}t | j|�std|� d���qt| jtd� | j||dd�}| jr.tj|dd�j }n|j391\}}}t� |d �� ||| f�j }t
| j�| _| j�|� | S )392a� Fit the spatial filters.393 394 Parameters395 ----------396 X : array, shape (n_epochs, n_channels, n_times)397 The data to be filtered.398 y : None | array, shape (n_samples,)399 Used for scikit-learn compatibility.400 401 Returns402 -------403 self : instance of UnsupervisedSpatialFilter404 Return the modified instance.405 )r rK rQ z6estimator must be a scikit-learn transformer, missing z methodr� Tr� r )Zaxis�r r r406 )r+ r� rE r r� �boolr1 r* r] �Tr, rU rT r �407estimator_r )r- r% r �attr�n_epochs�408n_channels�n_timesr& r& r0 r � s" ��� z UnsupervisedSpatialFilter.fitc C rj )a� Transform the data to its filtered components after fitting.409 410 Parameters411 ----------412 X : array, shape (n_epochs, n_channels, n_times)413 The data to be filtered.414 y : None | array, shape (n_samples,)415 Used for scikit-learn compatibility.416 417 Returns418 -------419 X : array, shape (n_epochs, n_channels, n_times)420 The transformed data.421 rP rM r& r& r0 rQ � s z'UnsupervisedSpatialFilter.fit_transformc C s t | j� | �|�}| �|d�S )a4 Transform the data to its spatial filters.422 423 Parameters424 ----------425 X : array, shape (n_epochs, n_channels, n_times)426 The data to be filtered.427 428 Returns429 -------430 X : array, shape (n_epochs, n_channels, n_times)431 The transformed data.432 rK )r r� r1 �
_apply_methodrJ r& r& r0 rK � s 433
434 z#UnsupervisedSpatialFilter.transformc C s | � |d�S )a= Inverse transform the data to its original space.435 436 Parameters437 ----------438 X : array, shape (n_epochs, n_components, n_times)439 The data to be inverted.440 441 Returns442 -------443 X : array, shape (n_epochs, n_channels, n_times)444 The transformed data.445 rN )r� rJ r& r& r0 rN � s
z+UnsupervisedSpatialFilter.inverse_transformc C sh |j \}}}t�|g d��}t�|||| g�j}t| j|�}||�}t�|jd||g��g d��}|S )aB Vectorize time samples as trials, apply method and reshape back.446 447 Parameters448 ----------449 X : array, shape (n_epochs, n_dims, n_times)450 The data to be inverted.451 452 Returns453 -------454 X : array, shape (n_epochs, n_dims, n_times)455 The transformed data.456 r� rR )r, r* rU rT r� �getattrr� )r- r% r� r� r� r� r& r& r0 r� � s
z'UnsupervisedSpatialFilter._apply_method)Fr7 )457r2 r3 r4 r5 r= r rQ rK rN r� r& r& r& r0 r� � s 458 459460$ r� c sR e Zd ZdZ ddd �Z� fd461d �Zdd d
�Zdd� Z� ZS )�TemporalFiltera� Estimator to filter data array along the last dimension.462 463 Applies a zero-phase low-pass, high-pass, band-pass, or band-stop464 filter to the channels.465 466 l_freq and h_freq are the frequencies below which and above which,467 respectively, to filter out of the data. Thus the uses are:468 469 - l_freq < h_freq: band-pass filter470 - l_freq > h_freq: band-stop filter471 - l_freq is not None, h_freq is None: low-pass filter472 - l_freq is None, h_freq is not None: high-pass filter473 474 See :func:`mne.filter.filter_data`.475 476 Parameters477 ----------478 l_freq : float | None479 Low cut-off frequency in Hz. If None the data are only low-passed.480 h_freq : float | None481 High cut-off frequency in Hz. If None the data are only482 high-passed.483 sfreq : float, default 1.0484 Sampling frequency in Hz.485 filter_length : str | int, default 'auto'486 Length of the FIR filter to use (if applicable):487 488 * int: specified length in samples.489 * 'auto' (default in 0.14): the filter length is chosen based490 on the size of the transition regions (7 times the reciprocal491 of the shortest transition band).492 * str: (default in 0.13 is "10s") a human-readable time in493 units of "s" or "ms" (e.g., "10s" or "5500ms") will be494 converted to that number of samples if ``phase="zero"``, or495 the shortest power-of-two length at least that duration for496 ``phase="zero-double"``.497 498 l_trans_bandwidth : float | str499 Width of the transition band at the low cut-off frequency in Hz500 (high pass or cutoff 1 in bandpass). Can be "auto"501 (default in 0.14) to use a multiple of ``l_freq``::502 503 min(max(l_freq * 0.25, 2), l_freq)504 505 Only used for ``method='fir'``.506 h_trans_bandwidth : float | str507 Width of the transition band at the high cut-off frequency in Hz508 (low pass or cutoff 2 in bandpass). Can be "auto"509 (default in 0.14) to use a multiple of ``h_freq``::510 511 min(max(h_freq * 0.25, 2.), info['sfreq'] / 2. - h_freq)512 513 Only used for ``method='fir'``.514 n_jobs : int | str, default 1515 Number of jobs to run in parallel.516 Can be 'cuda' if ``cupy`` is installed properly and method='fir'.517 method : str, default 'fir'518 'fir' will use overlap-add FIR filtering, 'iir' will use IIR519 forward-backward filtering (via filtfilt).520 iir_params : dict | None, default None521 Dictionary of parameters to use for IIR filtering.522 See mne.filter.construct_iir_filter for details. If iir_params523 is None and method="iir", 4th order Butterworth will be used.524 fir_window : str, default 'hamming'525 The window to use in FIR design, can be "hamming", "hann",526 or "blackman".527 fir_design : str528 Can be "firwin" (default) to use :func:`scipy.signal.firwin`,529 or "firwin2" to use :func:`scipy.signal.firwin2`. "firwin" uses530 a time-domain design technique that generally gives improved531 attenuation using fewer samples than "firwin2".532 533 .. versionadded:: 0.15534 535 See Also536 --------537 FilterEstimator538 Vectorizer539 mne.filter.filter_data540 NrC r� r� �hammingr� c C r� r7 ) r� r� rm r� r� r� rs r� r� �541fir_windowr� ) r- r� r� rm r� r� r� rs r� r� r� r� r& r& r0 r= Z r� zTemporalFilter.__init__c ru rv rw r{ r} r& r0 ry t r z TemporalFilter.__sklearn_tags__c C s d| _ | j||ddd� | S )a� Do nothing (for scikit-learn compatibility purposes).542 543 Parameters544 ----------545 X : array, shape ([n_epochs, ]n_channels, n_times)546 The data to be filtered over the last dimension. The channels547 dimension can be zero when passing a 2D array.548 y : None549 Not used, for scikit-learn compatibility issues.550 551 Returns552 -------553 self : instance of TemporalFilter554 The modified instance.555 TF)r r r )r� r1 rM r& r& r0 r { s zTemporalFilter.fitc C s� | j |dd�}t�|�}|jdkrtd|j� d���|j}|�d|d �}t|| j| j | j556| j | j | j
| j| j| jd| j| jd�
}|�|�S )an Filter data along the last dimension.557 558 Parameters559 ----------560 X : array, shape ([n_epochs, ]n_channels, n_times)561 The data to be filtered over the last dimension. The channels562 dimension can be zero when passing a 2D array.563 564 Returns565 -------566 X : array567 The data after filtering.568 Frd r\ z1Array must be of at max 3 dimensions instead got z dimensional matrixrR ) r� r� r� rs r� r� r r� r� )r1 r* Z569atleast_2dr_ rE r, rT r rm r� r� r� r� r� rs r� r� r� r� )r- r% r, r& r& r0 rK � s4 570571���572zTemporalFilter.transform) NNrC r� r� r� Nr� Nr� r� r7 ) r2 r3 r4 r5 r= ry r rK r� r& r& r} r0 r� s" S573� 574r� )+�numpyr* Z sklearn.baser r r r Zsklearn.preprocessingr r Z
sklearn.utilsr Zsklearn.utils.validationr Z575_fiff.pickr r r
r Zcovr Zepochsr �filterr Ztime_frequencyr �utilsr r r r Z_fixesr r r6 rX rY rh rk r� r� r� r& r& r& r0 �<module> s6 4# (hi ~