ENH: decoding module 2017

[kingjr]

This aims at keeping track of the development related to the Google summer of code 2016 on decoding analyses.

We have quite a lot ongoing PRs, so I thought I would try to organize them here to keep track of the big picture. I'll edit the post along the issue along the way.

The aim is to make transformers that follow the sklearn API:

pipe = make_pipeline(
    CSP(sfreq=200, None, 30),
    TimeFrequency(),
    SlidingEstimator(make_pipeline(StandardScaler(), LogisticRegression())
)

score = cross_val_score(X=epochs.get_data(), y=epochs.event[:, 2])

For now, we're focusing on sklearn integration, not high level features (plotting, get_coefs_ etc).

• Spatial filters:

  • FIX: XdawnTransformer: [MRG]: Xdawn Transformer compatible with sklearn #3245
  • FIX: Xdawn refactoring: MRG: Refactor Xdawn #3425
  • FIX: CSP renaming of epochs_data into X [MRG] multiclass CSP #3485
  • ENH: CSP adding multiclass support see ENH: CSP support for n_classes > 2 #3484, [MRG] multiclass CSP #3485
  • ENH: CSP uwedge multiclass + example ENH: CSP multiclass uwedge implementation + examples #3495
  • ENH: Unsupervised spatial filter [MRG] Spatial Filter #3447
  • ENH: EMS: ENH: EMS with sklearn object? #3427, MRG: EMS transformer with sklearn API #3446
  • ENH: CSP transform into power or time course or CSP transform doesn't comply with sklearn API #3583 MRG: add param to CSP to transform into average power or time course #3586
  • ENH: Mixin class to plot patterns_ and filters_? see BUG: time domain of the plot_pattern exceed range #3424
  • FIX: Move Xdawn to decoding?

• Freq - time/freq transformers:

  • FIX: TemporalFilter: needs refactoring from mne.decoding.FilterEstimator to pass explicit args and not info + BUG in FilterEstimator: BUG with FilterEstimator [DONT MERGE] #3395, ENH: Filterer for decoding #3471, [MRG+1] Created class Filterer and deprecated FilterEstimator #3472
  • FIX: PSDEstimator
  • ENH: TimeFrequencyDecomposer [MRG] TimeFrequency transformer for sklearn pipeline #3488
  • ENH: SPoC [MRG+1] Add SPoC spatial filtering for continuous target decoding #4144
  • Time Frequency Decoding object based on covariances / CSP ...
  • Power Decoding Estimator based on covariances / CSP: [MRG] Frequency CSP Example #4138 (example, but need to be converted in object)

• Continuous signals (raw not epochs)

  • FIX: Xdawn overlap: to be integrated with the RERP code ENH: refactor Xdawn and linear_regression_raw #2332, also see https://github.com/mne-tools/mne-python/blob/master/mne/stats/tests/test_regression.py#L124, PR in WIP: refactor XDawn to use rERP #3563
  • FIX: ReceptiveField: started in WIP: sklearn-style encoding / modularizing encoding pipelines #3310, done in [MRG+2] adding receptive field module #3728
  • FIX: combine rERP and STRF: e.g. float delays instead of list, discrete vs continuous regressors etc
  • Optimize ReceptiveField for continuous regression

• Search lights

  • ENH: SlidingEstimator and GeneralizingEstimator: MRG: Add search and generalization lights #3381 MRG+1: refactor time decoding & temporal generalization #4103
  • ENH: n-dimensional search light [MRG]: allow n_dimensional search lights #3481
  • FIX: Refactor TimeDecoding & GeneralizationAcrossTime MRG+1: refactor time decoding & temporal generalization #4103
  • ENH: 'scoring' param in SearchLight-like object ENH: specify scoring in search light #3475, MRG: Functionality to change scoring method of searchlight #3502, MRG: scoring in generalization light #3833 MRG+1: refactor time decoding & temporal generalization #4103
  • ENH: option to warm_start from one estimator to the next
  • cross_val_multiscore MRG+1: refactor time decoding & temporal generalization #4103
  • window_size and step parameter in SlidingEstimator

• Preprocessing

  • ENH: Scaler: scale channels over all time their corresponding time points to deal with issues related to mix channel types needs refactoring to comply to X y API
  • ENH: Vectorizer: to pass from n-D X to 2D X to be reviewed in [MRG] Vectorizer class to help chain MNE transformers by converting o/p into 2D #3409

• Examples:

  • RSA MRG: RSA example #3923
  • TimeFrequency Example MRG+1: Time Frequency CSP Example #4115
  • PSD example [MRG] Frequency CSP Example #4138
  • get_coef for SlidingEstimator
  • Denoise tutorial
  • Frequency generalization [WIP] Datasets from the BNCI database #4019
  • UnsupervisedSpatialFilter example

• Other:

  • partial_fit ENH: partial_fit for linear models #3483, WIP: partial_fit in search light #3591
  • get_coef(pipeline) to retrieve and/or invert_transform linear coefficients if they exist

[larsoner]

@kingjr is this issue still the roadmap, or should we close?

[kingjr]

Still rough roadmap, although needs to be updated. Let's close it after the sprint IMO.

[kingjr]

I'm actually for keeping it up ;)

[jona-sassenhagen]

@kingjr , @agramfort 's idea is one which will result in rERP and the receptive field module staying separate forever.

• rERP becomes restricted to the original idea, allowing only sparse predictors; dense predictors are handled by receptive field module
• receptive field module should be extended slightly to simply also using sparse predictors
  (- XDawn and rERP should still be united)

[jona-sassenhagen]

I closed/crossed out a few that had been done since.
8 issues still open.

[jona-sassenhagen]

@kingjr any idea for an example for UnsupervisedSpatialFilter with the example data? The only thing I can think of is show that you can use an unregularized algorithm when p > n by reducing p in dimensionality via PCA, i.e., for the MEG data with many channels and few trials.

[kingjr]

How about PCA and then sliding with svm multiclass on the left/right audio/visual dataset ?

…

On Saturday, 4 August 2018, jona-sassenhagen ***@***.***> wrote: @kingjr <https://github.com/kingjr> any idea for an example for UnsupervisedSpatialFilter with the example data? The only thing I can think of is show that you can use an unregularized algorithm when p > n by reducing p in dimensionality via PCA, i.e., for the MEG data with many channels and few trials. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3442 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEp7DPMjiTgyOOWCHNAnB7cb8KjPACRaks5uNgcxgaJpZM4JRF5A> .