Models Summary#
This page offers a summary of all braindecode implemented models. For more information on each model, please consult the API.
Columns definitions:
Model: The name of the model.
Application: The application(s) the model is typically used for (e.g., Motor Imagery, P300, Sleep Staging). ‘General’ indicates applicability across multiple applications or no specific application focus.
Type: The model’s primary function (e.g., Classification, Regression, Embedding).
Sampling Frequency: The data sampling rate (in Hertz) the model is designed for. Note that this might be adaptable depending on the specific dataset and application.
Categorization: models categorization based on the main building blocks used in the architecture. See Models Categorization page for more details.
- Hyperparameters: The mandatory hyperparameters required for instantiating the model class. These may include:
n_chans, number of channels/electrodes/sensors,
n_outputs, number of output classes or regression targets,
n_times, number of time points in the input window,
freq (Hz), sampling frequency,
chs_info, information about each individual EEG channel. Refer to
mne.Info(see its “chs” field for details)Also, n_times can be derived implicitly by providing both sfreq and input_window_seconds.
#Parameters: The approximate total number of trainable parameters in the model, calculated using a consistent configuration (see note below).
| Model | Application | Type | Categorization | Sampling Frequency (Hz) | #Parameters | Hyperparameters |
|---|---|---|---|---|---|---|
| ATCNet | General | Classification | Convolution Recurrent Small Attention | 250 | 113732 | n_chans n_outputs n_times |
| AttentionBaseNet | Motor Imagery | Classification | Convolution Small Attention | 250 | 3692 | n_chans n_outputs n_times |
| BDTCN | Normal Abnormal | Classification | Convolution Recurrent | 100 | 456502 | n_chans n_outputs n_times |
| BIOT | Sleep Staging Epilepsy | Classification | Large Language Model | 200 | 3183879 | n_chans n_outputs |
| ContraWR | Sleep Staging | Classification Embedding | Convolution | 125 | 1160165 | n_chans n_outputs freq (Hz) |
| CTNet | Motor Imagery | Classification | Convolution Small Attention | 250 | 26900 | n_chans n_outputs n_times |
| Deep4Net | General | Classification | Convolution | 250 | 282879 | n_chans n_outputs n_times |
| DeepSleepNet | Sleep Staging | Classification | Convolution Recurrent | 256 | 24744837 | n_chans n_outputs |
| EEGConformer | General | Classification | Convolution Small Attention | 250 | 789572 | n_chans n_outputs n_times |
| EEGInceptionERP | ERP SSVEP | Classification | Convolution | 128 | 14926 | n_chans n_outputs |
| EEGInceptionMI | Motor Imagery | Classification | Convolution | 250 | 558028 | n_chans n_outputs n_times |
| EEGITNet | Motor Imagery | Classification | Convolution Recurrent | 125 | 5212 | n_chans n_outputs n_times |
| EEGNet | General | Classification | Convolution | 128 | 2484 | n_chans n_outputs n_times |
| EEGNeX | Motor Imagery | Classification | Convolution | 125 | 55940 | n_chans n_outputs n_times |
| EEGMiner | Emotion Recognition | Classification | Convolution Interpretability | 128 | 7572 | n_chans n_outputs n_times freq (Hz) |
| EEGSimpleConv | Motor Imagery | Classification | Convolution | 80 | 730404 | n_chans n_outputs freq (Hz) |
| EEGTCNet | Motor Imagery | Classification | Convolution Recurrent | 250 | 4516 | n_chans n_outputs |
| Labram | General | Classification Embedding | Convolution Large Language Model | 200 | 5866180 | n_chans n_outputs n_times |
| MSVTNet | Motor Imagery | Classification | Convolution Recurrent Small Attention | 250 | 75494 | n_chans n_outputs n_times |
| SCCNet | Motor Imagery | Classification | Convolution | 125 | 12070 | n_chans n_outputs n_times freq (Hz) |
| SignalJEPA | Motor Imagery ERP SSVEP | Embedding | Convolution Channel Large Language Model | 128 | 3456882 | n_times chs_info |
| SignalJEPA_Contextual | Motor Imagery ERP SSVEP | Classification | Convolution Channel Large Language Model | 128 | 3459184 | n_outputs n_times chs_info |
| SignalJEPA_PostLocal | Motor Imagery ERP SSVEP | Classification | Convolution Channel Large Language Model | 128 | 16142 | n_chans n_outputs n_times |
| SignalJEPA_PreLocal | Motor Imagery ERP SSVEP | Classification | Convolution Channel Large Language Model | 128 | 16142 | n_outputs n_times chs_info |
| SincShallowNet | Motor Imagery | Classification | Convolution Interpretability | 250 | 21892 | n_chans n_outputs n_times freq (Hz) |
| ShallowFBCSPNet | General | Classification | Convolution | 250 | 46084 | n_chans n_outputs n_times |
| SleepStagerBlanco2020 | Sleep Staging | Classification | Convolution | 100 | 2845 | n_chans n_outputs n_times |
| SleepStagerChambon2018 | Sleep Staging | Classification | Convolution | 128 | 5835 | n_chans n_outputs n_times freq (Hz) |
| AttnSleep | Sleep Staging | Classification | Convolution Small Attention | 100 | 719925 | n_chans n_outputs n_times freq (Hz) |
| SPARCNet | Epilepsy | Classification | Convolution | 200 | 1141921 | n_chans n_outputs n_times |
| SyncNet | Emotion Recognition Alcoholism | Classification | Interpretability | 256 | 554 | n_chans n_outputs n_times |
| TSception | Emotion Recognition | Classification | Convolution | 256 | 2187206 | n_chans n_outputs n_times freq (Hz) |
| TIDNet | General | Classification | Convolution | 250 | 240404 | n_chans n_outputs n_times |
| USleep | Sleep Staging | Classification | Convolution | 128 | 2482011 | n_chans n_outputs n_times freq (Hz) |
| FBCNet | Motor Imagery | Classification | Convolution FilterBank | 250 | 11812 | n_chans n_outputs n_times freq (Hz) |
| FBMSNet | Motor Imagery | Classification | Convolution FilterBank | 250 | 16231 | n_chans n_outputs n_times freq (Hz) |
| FBLightConvNet | Motor Imagery | Classification | Convolution FilterBank | 250 | 6596 | n_chans n_outputs n_times freq (Hz) |
| IFNet | Motor Imagery | Classification | Convolution FilterBank | 250 | 9860 | n_chans n_outputs n_times freq (Hz) |
The parameter counts shown in the table were calculated using consistent hyperparameters for models within the same paradigm, based largely on Braindecode’s default implementation values. These counts provide a relative comparison but may differ from those reported in the original publications due to variations in specific architectural details, input dimensions used in the paper, or calculation methods.
We are continually expanding this collection and welcome contributions! If you have implemented a model relevant to EEG, EcoG, or MEG analysis, consider adding it to Braindecode.