obob_mne.decoding.GeneralizedTemporal

class obob_mne.decoding.GeneralizedTemporal(epochs, pipeline, epochs_test=None, cv=2, scoring='accuracy', n_jobs=-1, metadata_querylist=None, metadata_querylist_test=None)

Apply a decoding pipeline using Temporal Generalization.

Use this class to perform Temporal Generalization decoding. This means that a classifier is trained on every sample and then tested on all samples. So, if you supply epochs with 100 samples, a classifier gets trained on the data data of the first sample. This classifier is then tested on the data of the first sample, then the second sample and so on. The process is then repeated by training on the second sample and testing on all samples and so forth.

The minimum requirements are the data as mne.Epochs and the pipeline as sklearn.pipeline.Pipeline, which is most commonly generated using sklearn.pipeline.make_pipeline().

This class treats every individual event_id as an individual target. If you want to combine multiple events_id into one target, you can use epochs.collapse_conditions().

If you want the training and testing data to be different, you can supply the training data to epochs and the testing data to epochs_test.

Cross validation will be run only if training and testing data are equal (i.e., when epochs_test=None).

Parameters:

• epochs (mne.Epochs) – The epochs to apply the classifier pipeline to.
• pipeline (sklearn.pipeline.Pipeline) – The classifier pipeline to use. Most likely created with sklearn.pipeline.make_pipeline()
• epochs_test (mne.Epochs or None, optional) – If set, the classifier gets tested on this data. The event_ids must be equal. No crossvalidation will be run in this case.
• cv (int, optional) – The amount of folds for cross validation
• scoring (str or callable, optional) – The scoring function to use.
• n_jobs (int, optional) – Number of CPU cores to use