Braindecode 1.2.0.dev175267687__tar.gz → 1.2.0.dev175337561__tar.gz

{braindecode-1.2.0.dev175267687/braindecode.egg-info → braindecode-1.2.0.dev175337561}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: braindecode
-Version: 1.2.0.dev175267687
+Version: 1.2.0.dev175337561
 Summary: Deep learning software to decode EEG, ECG or MEG signals
 Author-email: Robin Tibor Schirrmeister <robintibor@gmail.com>
 Maintainer-email: Alexandre Gramfort <agramfort@meta.com>, Bruno Aristimunha Pinto <b.aristimunha@gmail.com>, Robin Tibor Schirrmeister <robintibor@gmail.com>
@@ -17,7 +17,7 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Python: >=3.10
+Requires-Python: >3.10
 Description-Content-Type: text/x-rst
 License-File: LICENSE.txt
 License-File: NOTICE.txt
@@ -49,10 +49,6 @@ Requires-Dist: mypy; extra == "tests"
 Provides-Extra: docs
 Requires-Dist: sphinx_gallery; extra == "docs"
 Requires-Dist: sphinx_rtd_theme; extra == "docs"
-Requires-Dist: sphinx-autodoc-typehints; extra == "docs"
-Requires-Dist: sphinx-autobuild; extra == "docs"
-Requires-Dist: sphinxcontrib-bibtex; extra == "docs"
-Requires-Dist: sphinx_sitemap; extra == "docs"
 Requires-Dist: pydata_sphinx_theme; extra == "docs"
 Requires-Dist: numpydoc; extra == "docs"
 Requires-Dist: memory_profiler; extra == "docs"
@@ -63,7 +59,6 @@ Requires-Dist: lightning; extra == "docs"
 Requires-Dist: seaborn; extra == "docs"
 Requires-Dist: pre-commit; extra == "docs"
 Requires-Dist: openneuro-py; extra == "docs"
-Requires-Dist: plotly; extra == "docs"
 Provides-Extra: all
 Requires-Dist: braindecode[docs,moabb,tests]; extra == "all"
 Dynamic: license-file

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/__init__.py

@@ -4,7 +4,6 @@ Some predefined network architectures for EEG decoding.
 
 from .atcnet import ATCNet
 from .attentionbasenet import AttentionBaseNet
-from .attn_sleep import AttnSleep
 from .base import EEGModuleMixin
 from .biot import BIOT
 from .contrawr import ContraWR
@@ -16,8 +15,9 @@ from .eeginception_erp import EEGInceptionERP
 from .eeginception_mi import EEGInceptionMI
 from .eegitnet import EEGITNet
 from .eegminer import EEGMiner
-from .eegnet import EEGNet, EEGNetv4
+from .eegnet import EEGNetv1, EEGNetv4
 from .eegnex import EEGNeX
+from .eegresnet import EEGResNet
 from .eegsimpleconv import EEGSimpleConv
 from .eegtcnet import EEGTCNet
 from .fbcnet import FBCNet
@@ -38,11 +38,12 @@ from .signal_jepa import (
 from .sinc_shallow import SincShallowNet
 from .sleep_stager_blanco_2020 import SleepStagerBlanco2020
 from .sleep_stager_chambon_2018 import SleepStagerChambon2018
+from .sleep_stager_eldele_2021 import SleepStagerEldele2021
 from .sparcnet import SPARCNet
 from .syncnet import SyncNet
 from .tcn import BDTCN, TCN
 from .tidnet import TIDNet
-from .tsinception import TSception
+from .tsinception import TSceptionV1
 from .usleep import USleep
 from .util import _init_models_dict, models_mandatory_parameters
 
@@ -52,7 +53,6 @@ _init_models_dict()
 
 __all__ = [
     "ATCNet",
-    "AttnSleep",
     "AttentionBaseNet",
     "EEGModuleMixin",
     "BIOT",
@@ -65,9 +65,10 @@ __all__ = [
     "EEGInceptionMI",
     "EEGITNet",
     "EEGMiner",
-    "EEGNet",
+    "EEGNetv1",
     "EEGNetv4",
     "EEGNeX",
+    "EEGResNet",
     "EEGSimpleConv",
     "EEGTCNet",
     "FBCNet",
@@ -86,12 +87,13 @@ __all__ = [
     "SincShallowNet",
     "SleepStagerBlanco2020",
     "SleepStagerChambon2018",
+    "SleepStagerEldele2021",
     "SPARCNet",
     "SyncNet",
     "BDTCN",
     "TCN",
     "TIDNet",
-    "TSception",
+    "TSceptionV1",
     "USleep",
     "_init_models_dict",
     "models_mandatory_parameters",

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/atcnet.py

@@ -138,7 +138,7 @@ class ATCNet(EEGModuleMixin, nn.Module):
     .. rubric:: Usage and Configuration
 
         - ``conv_block_n_filters (F1)``, ``conv_block_depth_mult (D)`` → capacity of the stem
-        (with ``F2 = F1·D`` feeding attention/TCN), dimensions aligned to ``F2``, like :class:`EEGNet`.
+        (with ``F2 = F1·D`` feeding attention/TCN), dimensions aligned to ``F2``, like `EEGNetv4`.
         - Pool sizes ``P1,P2`` trade temporal resolution for stability/compute; they set
         ``T_c = T/(P1·P2)`` and thus window width ``T_w``.
         - ``n_windows`` controls the ensemble over shifts (compute ∝ windows).

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/attentionbasenet.py

@@ -24,7 +24,7 @@ from braindecode.modules.attention import (
 
 
 class AttentionBaseNet(EEGModuleMixin, nn.Module):
-    """AttentionBaseNet from Wimpff M et al. (2023) [Martin2023]_.
+    """
 
     :bdg-success:`Convolution` :bdg-info:`Small Attention`
 
@@ -57,9 +57,9 @@ class AttentionBaseNet(EEGModuleMixin, nn.Module):
 
         - *Operations.*
         - **Temporal conv** (:class:`torch.nn.Conv2d`) with kernel ``(1, L_t)`` creates a learned
-          FIR-like filter bank with ``n_temporal_filters`` maps.
+        FIR-like filter bank with ``n_temporal_filters`` maps.
         - **Depthwise spatial conv** (:class:`torch.nn.Conv2d`, ``groups=n_temporal_filters``)
-          with kernel ``(n_chans, 1)`` learns per-filter spatial projections over the full montage.
+        with kernel ``(n_chans, 1)`` learns per-filter spatial projections over the full montage.
         - **BatchNorm → ELU → AvgPool → Dropout** stabilize and downsample time.
         - Output shape: ``(B, F2, 1, T₁)`` with ``F2 = n_temporal_filters x spatial_expansion``.
 
@@ -71,8 +71,8 @@ class AttentionBaseNet(EEGModuleMixin, nn.Module):
 
         - *Operations.*
         - A ``1x1`` conv → BN → activation maps ``F2 → ch_dim`` without changing
-          the temporal length ``T₁`` (shape: ``(B, ch_dim, 1, T₁)``).
-          This sets the embedding width for the attention block.
+        the temporal length ``T₁`` (shape: ``(B, ch_dim, 1, T₁)``).
+        This sets the embedding width for the attention block.
 
     - :class:`_ChannelAttentionBlock` **(temporal refinement + channel attention)**
 
@@ -117,10 +117,8 @@ class AttentionBaseNet(EEGModuleMixin, nn.Module):
     - **Type.** Channel attention chosen by ``attention_mode`` (SE, ECA, CBAM, CAT, GSoP,
         EncNet, GE, GCT, SRM, CATLite). Most operate purely on channels; CBAM/CAT additionally
         include temporal attention.
-
     - **Shapes.** Input/Output around attention: ``(B, ch_dim, 1, T₁)``. Re-arrangements
         (if any) are internal to the module; the block returns the same shape before pooling.
-
     - **Role.** Re-weights channels (and optionally time) to highlight informative sources
         and suppress distractors, improving SNR ahead of the linear head.
 
@@ -165,11 +163,10 @@ class AttentionBaseNet(EEGModuleMixin, nn.Module):
     Notes
     -----
     - Sequence length after each stage is computed internally; the final classifier expects
-      a flattened ``ch_dim x T₂`` vector.
+    a flattened ``ch_dim x T₂`` vector.
     - Attention operates on *channel* dimension by design; temporal gating exists only in
-      specific variants (CBAM/CAT).
-    - The paper and original code with more details about the methodological
-      choices are available at the [Martin2023]_ and [MartinCode]_.
+    specific variants (CBAM/CAT).
+
     .. versionadded:: 0.9
 
     Parameters
@@ -198,18 +195,18 @@ class AttentionBaseNet(EEGModuleMixin, nn.Module):
         the depth of the network after the initial layer. Default is 16.
     attention_mode : str, optional
         The type of attention mechanism to apply. If `None`, no attention is applied.
-            - "se" for Squeeze-and-excitation network
-            - "gsop" for Global Second-Order Pooling
-            - "fca" for Frequency Channel Attention Network
-            - "encnet" for context encoding module
-            - "eca" for Efficient channel attention for deep convolutional neural networks
-            - "ge" for Gather-Excite
-            - "gct" for Gated Channel Transformation
-            - "srm" for Style-based Recalibration Module
-            - "cbam" for Convolutional Block Attention Module
-            - "cat" for Learning to collaborate channel and temporal attention
-            from multi-information fusion
-            - "catlite" for Learning to collaborate channel attention
+        - "se" for Squeeze-and-excitation network
+        - "gsop" for Global Second-Order Pooling
+        - "fca" for Frequency Channel Attention Network
+        - "encnet" for context encoding module
+        - "eca" for Efficient channel attention for deep convolutional neural networks
+        - "ge" for Gather-Excite
+        - "gct" for Gated Channel Transformation
+        - "srm" for Style-based Recalibration Module
+        - "cbam" for Convolutional Block Attention Module
+        - "cat" for Learning to collaborate channel and temporal attention
+        from multi-information fusion
+        - "catlite" for Learning to collaborate channel attention
         from multi-information fusion (lite version, cat w/o temporal attention)
     pool_length : int, default=8
         The length of the window for the average pooling operation.

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/ctnet.py

@@ -39,7 +39,7 @@ class CTNet(EEGModuleMixin, nn.Module):
     The architecture consists of three main components:
 
     1. **Convolutional Module**:
-        - Apply :class:`EEGNet` to perform some feature extraction, denoted here as
+        - Apply EEGNetV4 to perform some feature extraction, denoted here as
         _PatchEmbeddingEEGNet module.
 
     2. **Transformer Encoder Module**:

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/deep4.py

@@ -19,13 +19,9 @@ from braindecode.modules import (
 class Deep4Net(EEGModuleMixin, nn.Sequential):
     """Deep ConvNet model from Schirrmeister et al (2017) [Schirrmeister2017]_.
 
-    :bdg-success:`Convolution`
-
-    .. figure:: https://onlinelibrary.wiley.com/cms/asset/fc200ccc-d8c4-45b4-8577-56ce4d15999a/hbm23730-fig-0001-m.jpg
+     .. figure:: https://onlinelibrary.wiley.com/cms/asset/fc200ccc-d8c4-45b4-8577-56ce4d15999a/hbm23730-fig-0001-m.jpg
         :align: center
-        :alt: Deep4Net Architecture
-        :width: 600px
-
+        :alt: CTNet Architecture
 
     Model described in [Schirrmeister2017]_.
 

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/deepsleepnet.py

@@ -8,128 +8,14 @@ from braindecode.models.base import EEGModuleMixin
 
 
 class DeepSleepNet(EEGModuleMixin, nn.Module):
-    """DeepSleepNet from Supratak et al. (2017) [Supratak2017]_.
+    """Sleep staging architecture from Supratak et al. (2017) [Supratak2017]_.
 
-    :bdg-success:`Convolution` :bdg-info:`Recurrent`
-
-    .. figure:: https://raw.githubusercontent.com/akaraspt/deepsleepnet/master/img/deepsleepnet.png
+     .. figure:: https://raw.githubusercontent.com/akaraspt/deepsleepnet/refs/heads/master/img/deepsleepnet.png
         :align: center
         :alt: DeepSleepNet Architecture
-        :width: 700px
-
-    .. rubric:: Architectural Overview
-
-    DeepSleepNet couples **dual-path convolution neural network representation learning** with
-    **sequence residual learning** via bidirectional LSTMs.
-
-    The network have:
-
-    - (i) learns complementary, time-frequency features from each
-      30-s epoch using **two parallel CNNs** (small vs. large first-layer filters), then
-    - (ii) models **temporal dependencies across epochs** using **two-layer BiLSTMs**
-      with a **residual shortcut** from the CNN features, and finally
-    - (iii) outputs per-epoch sleep stages. This design encodes both
-      epoch-local patterns and longer-range transition rules used by human scorers.
-
-    In term of implementation:
-
-    - (i) :class:`_RepresentationLearning` two CNNs extract epoch-wise features
-      (small-filter path for temporal precision; large-filter path for frequency precision);
-    - (ii) :class:`_SequenceResidualLearning` stacked BiLSTMs with peepholes + residual shortcut
-      inject temporal context while preserving CNN evidence;
-    - (iii) :class:`_Classifier` linear readout (softmax) for the five sleep stages.
-
-    .. rubric:: Macro Components
-
-    - :class:`_RepresentationLearning` **(dual-path CNN → epoch feature)**
-
-       - *Operations.*
-       - **Small-filter CNN** 4 times:
-       - :class:`~torch.nn.Conv1d`
-       - :class:`~torch.nn.BatchNorm1d`
-       - :class:`~torch.nn.ReLU`
-       - :class:`~torch.nn.MaxPool1d` after.
-       First conv uses **filter length ≈ Fs/2** and **stride ≈ Fs/16** to emphasize *timing* of graphoelements.
-    - **Large-filter CNN**:
-        - Same stack but first conv uses **filter length ≈ 4·Fs** and
-        - **stride ≈ Fs/2** to emphasize *frequency* content.
-    - Outputs from both paths are **concatenated** into the epoch embedding ``a_t``.
-
-    - *Rationale.*
-      Two first-layer scales provide a **learned, dual-scale filter bank** that trades
-      temporal vs. frequency precision without hand-crafted features.
-
-    - :class:`_SequenceResidualLearning` (:class:`~torch.nn.BiLSTM` **context + residual fusion)**
-
-        - *Operations.*
-        - **Two-layer BiLSTM** with **peephole connections** processes the sequence of epoch embeddings
-          ``{a_t}`` forward and backward; hidden states from both directions are **concatenated**.
-        - A **shortcut MLP** (fully connected + :class:`~torch.nn.BatchNorm1d` + :class:`~torch.nn.ReLU`) projects ``a_t`` to the BiLSTM output
-          dimension and is **added** (residual) to the :class:`~torch.nn.BiLSTM` output at each time step.
-        - *Role.* Encodes **stage-transition rules** and smooths predictions over time while preserving
-          salient CNN features via the residual path.
-
-    - :class:`_Classifier` **(epoch-wise prediction)**
-
-        - *Operations.*
-        - :class:`~torch.nn.Linear` to produce per-epoch class probabilities.
 
-    Original training uses two-step optimization: CNN pretraining on class-balanced data,
-    then end-to-end fine-tuning with sequential batches.
-
-    .. rubric:: Convolutional Details
-
-    - **Temporal (where time-domain patterns are learned).**
-
-    Both CNN paths use **1-D temporal convolutions**. The *small-filter* path (first kernel ≈ Fs/2,
-    stride ≈ Fs/16) captures *when* characteristic transients occur; the *large-filter* path
-    (first kernel ≈ 4·Fs, stride ≈ Fs/2) captures *which* frequency components dominate over the
-    epoch. Deeper layers use **small kernels** to refine features with fewer parameters, interleaved
-    with **max pooling** for downsampling.
-
-    - **Spatial (how channels are processed).**
-    The original model operates on **single-channel** raw EEG; convolutions therefore mix only
-    along time (no spatial convolution across electrodes).
-
-    - **Spectral (how frequency information emerges).**
-    No explicit Fourier/wavelet transform is used. The **large-filter path** serves as a
-    *frequency-sensitive* analyzer, while the **small-filter path** remains *time-sensitive*,
-    together functioning as a **two-band learned filter bank** at the first layer.
-
-    .. rubric:: Attention / Sequential Modules
-
-    - **Type.** **Bidirectional LSTM** (two layers) with **peephole connections**; forward and
-      backward streams are independent and concatenated.
-    - **Shapes.** For a sequence of ``N`` epochs, the CNN produces ``{a_t} ∈ R^{D}``;
-      BiLSTM outputs ``h_t ∈ R^{2H}``; the shortcut MLP maps ``a_t → R^{2H}`` to enable
-      **element-wise residual addition**.
-    - **Role.** Models **long-range temporal dependencies** (e.g., persisting N2 without visible
-      K-complex/spindles), stabilizing per-epoch predictions.
-
-
-    .. rubric:: Additional Mechanisms
-
-    - **Residual shortcut over sequence encoder.** Adds projected CNN features to BiLSTM outputs,
-      improving gradient flow and retaining discriminative content from representation learning.
-    - **Two-step training.**
-        - (i) **Pretrain** the CNN paths with class-balanced sampling;
-        - (ii) **fine-tune** the full network with sequential batches, using **lower LR** for CNNs and **higher LR** for the
-        sequence encoder.
-    - **State handling.** BiLSTM states are **reinitialized per subject** so that temporal context
-      does not leak across recordings.
-
-
-    .. rubric:: Usage and Configuration
-
-    - **Epoch pipeline.** Use **two parallel CNNs** with the first conv sized to **Fs/2** (small path)
-      and **4·Fs** (large path), with strides **Fs/16** and **Fs/2**, respectively; stack three more
-      conv blocks with small kernels, plus **max pooling** in each path. Concatenate path outputs
-      to form epoch embeddings.
-    - **Sequence encoder.** Apply **two-layer BiLSTM (peepholes)** over the sequence of embeddings;
-      add a **projection MLP** on the CNN features and **sum** with BiLSTM outputs (residual).
-      Finish with :class:`~torch.nn.Linear` per epoch.
-    - **Reference implementation.** See the official repository for a faithful implementation and
-      training scripts.
+    Convolutional neural network and bidirectional-Long Short-Term
+    for single channels sleep staging described in [Supratak2017]_.
 
     Parameters
     ----------
@@ -146,6 +32,7 @@ class DeepSleepNet(EEGModuleMixin, nn.Module):
     drop_prob : float, default=0.5
         The dropout rate for regularization. Values should be between 0 and 1.
 
+
     References
     ----------
     .. [Supratak2017] Supratak, A., Dong, H., Wu, C., & Guo, Y. (2017).

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/eegconformer.py

@@ -27,25 +27,22 @@ class EEGConformer(EEGModuleMixin, nn.Module):
     EEG-Conformer is a *convolution-first* model augmented with a *lightweight transformer
     encoder*. The end-to-end flow is:
 
-    - (i) :class:`_PatchEmbedding` converts the continuous EEG into a compact sequence of tokens via a
-      :class:`ShallowFBCSPNet` temporal–spatial conv stem and temporal pooling;
-    - (ii) :class:`_TransformerEncoder` applies small multi-head self-attention to integrate
-      longer-range temporal context across tokens;
+    - (i) :class:`_PatchEmbedding` converts the continuous EEG into a compact sequence of tokens via a :class:`ShallowFBCSPNet` temporal–spatial conv stem and temporal pooling;
+    - (ii) :class:`_TransformerEncoder applies small multi-head self-attention to integrate longer-range temporal context across tokens;
     - (iii) :class:`_ClassificationHead` aggregates the sequence and performs a linear readout.
-      This preserves the strong inductive biases of shallow CNN filter banks while adding
-      just enough attention to capture dependencies beyond the pooling horizon [song2022]_.
+        This preserves the strong inductive biases of shallow CNN filter banks while adding
+        just enough attention to capture dependencies beyond the pooling horizon [song2022]_.
 
     .. rubric:: Macro Components
 
     - :class:`_PatchEmbedding` **(Shallow conv stem → tokens)**
 
-        - *Operations.*
-        - A temporal convolution (`:class:torch.nn.Conv2d`) ``(1 x L_t)`` forms a data-driven "filter bank";
-        - A spatial convolution (`:class:torch.nn.Conv2d`) (n_chans x 1)`` projects across electrodes,
-          collapsing the channel axis into a virtual channel.
-        - **Normalization function** :class:`torch.nn.BatchNorm`
-        - **Activation function** :class:`torch.nn.ELU`
-        - **Average Pooling** :class:`torch.nn.AvgPool` along time (kernel ``(1, P)`` with stride ``(1, S)``)
+    - *Operations.*
+        - A temporal convolution (`:class:`torch.nn.Conv2d`) ``(1 x L_t)`` forms a data-driven "filter bank";
+        - A spatial convolution (`:class:`torch.nn.Conv2d`) (n_chans x 1)`` projects across electrodes, collapsing the channel axis into a virtual channel.
+        - **Normalization function** `:class:torch.nn.BatchNorm`
+        - **Activation function** `:class:torch.nn.ELU`
+        - **Average Pooling** `:class:torch.nn.AvgPool` along time (kernel ``(1, P)`` with stride ``(1, S)``)
         -  final ``1x1`` :class:`torch.nn.Linear` projection.
 
     The result is rearranged to a token sequence ``(B, S_tokens, D)``, where ``D = n_filters_time``.
@@ -56,7 +53,7 @@ class EEGConformer(EEGModuleMixin, nn.Module):
 
     - :class:`_TransformerEncoder` **(context over temporal tokens)**
 
-        - *Operations.*
+    - *Operations.*
         - A stack of ``att_depth`` encoder blocks. :class:`_TransformerEncoderBlock`
         - Each block applies LayerNorm :class:`torch.nn.LayerNorm`
         - Multi-Head Self-Attention (``att_heads``) with dropout + residual :class:`MultiHeadAttention` (:class:`torch.nn.Dropout`)
@@ -70,7 +67,7 @@ class EEGConformer(EEGModuleMixin, nn.Module):
 
     - :class:`ClassificationHead` **(aggregation + readout)**
 
-        - *Operations*.
+    - *Operations*.
         - Flatten, :class:`torch.nn.Flatten` the sequence ``(B, S_tokens·D)`` -
         - MLP (:class:`torch.nn.Linear` → activation (default: :class:`torch.nn.ELU`) → :class:`torch.nn.Dropout` → :class:`torch.nn.Linear`)
         - final Linear to classes.

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/eeginception_erp.py

@@ -15,84 +15,12 @@ from braindecode.modules import DepthwiseConv2d, Ensure4d, InceptionBlock
 class EEGInceptionERP(EEGModuleMixin, nn.Sequential):
     """EEG Inception for ERP-based from Santamaria-Vazquez et al (2020) [santamaria2020]_.
 
-    :bdg-success:`Convolution`
-
     .. figure:: https://braindecode.org/dev/_static/model/eeginceptionerp.jpg
         :align: center
         :alt: EEGInceptionERP Architecture
 
-        Figure: Overview of EEG-Inception architecture. 2D convolution blocks and depthwise 2D convolution blocks include batch normalization, activation and dropout regularization. The kernel size is displayed for convolutional and average pooling layers.
-
-    .. rubric:: Architectural Overview
-
-    A two-stage, multi-scale CNN tailored to ERP detection from short (0-1000 ms) single-trial epochs. Signals are mapped through
-    * (i) :class:`_InceptionModule1` multi-scale temporal feature extraction plus per-branch spatial mixing;
-    * (ii) :class:`_InceptionModule2` deeper multi-scale refinement at a reduced temporal resolution; and
-    * (iii) :class:`_OutputModule` compact aggregation and linear readout.
-
-    .. rubric:: Macro Components
-
-    - :class:`_InceptionModule1` **(multi-scale temporal + spatial mixing)**
-
-        - *Operations.*
-        - `EEGInceptionERP.c1`: :class:`torch.nn.Conv2d` ``k=(64,1)``, stride ``(1,1)``, *same* pad on input reshaped to ``(B,1,128,8)`` → BN → activation → dropout.
-        - `EEGInceptionERP.d1`: :class:`torch.nn.Conv2d` (depthwise) ``k=(1,8)``, *valid* pad over channels → BN → activation → dropout.
-        - `EEGInceptionERP.c2`: :class:`torch.nn.Conv2d` ``k=(32,1)`` → BN → activation → dropout; then `EEGInceptionERP.d2` depthwise ``k=(1,8)`` → BN → activation → dropout.
-        - `EEGInceptionERP.c3`: :class:`torch.nn.Conv2d` ``k=(16,1)`` → BN → activation → dropout; then `EEGInceptionERP.d3` depthwise ``k=(1,8)`` → BN → activation → dropout.
-        - `EEGInceptionERP.n1`: :class:`torch.nn.Concat` over branch features.
-        - `EEGInceptionERP.a1`: :class:`torch.nn.AvgPool2d` ``pool=(4,1)``, stride ``(4,1)`` for temporal downsampling.
-
-    *Interpretability/robustness.* Depthwise `1 x n_chans` layers act as learnable montage-wide spatial filters per temporal scale; pooling stabilizes against jitter.
-
-    - :class:`_InceptionModule2` **(refinement at coarser timebase)**
-
-        - *Operations.*
-        - `EEGInceptionERP.c4`: :class:`torch.nn.Conv2d` ``k=(16,1)`` → BN → activation → dropout.
-        - `EEGInceptionERP.c5`: :class:`torch.nn.Conv2d` ``k=(8,1)`` → BN → activation → dropout.
-        - `EEGInceptionERP.c6`: :class:`torch.nn.Conv2d` ``k=(4,1)`` → BN → activation → dropout.
-        - `EEGInceptionERP.n2`: :class:`torch.nn.Concat` (merge C4-C6 outputs).
-        - `EEGInceptionERP.a2`: :class:`torch.nn.AvgPool2d` ``pool=(2,1)``, stride ``(2,1)``.
-        - `EEGInceptionERP.c7`: :class:`torch.nn.Conv2d` ``k=(8,1)`` → BN → activation → dropout; then `EEGInceptionERP.a3`: :class:`torch.nn.AvgPool2d` ``pool=(2,1)``.
-        - `EEGInceptionERP.c8`: :class:`torch.nn.Conv2d` ``k=(4,1)`` → BN → activation → dropout; then `EEGInceptionERP.a4`: :class:`torch.nn.AvgPool2d` ``pool=(2,1)``.
-
-    *Role.* Adds higher-level, shorter-window evidence while progressively compressing temporal dimension.
-
-    - :class:`_OutputModule` **(aggregation + readout)**
-
-        - *Operations.*
-        - :class:`torch.nn.Flatten`
-        - :class:`torch.nn.Linear` ``(features → 2)``
-
-    .. rubric:: Convolutional Details
-
-    - **Temporal (where time-domain patterns are learned).**
-    First module uses 1D temporal kernels along the 128-sample axis: ``64``, ``32``, ``16``
-    (≈500, 250, 125 ms at 128 Hz). After ``pool=(4,1)``, the second module applies ``16``,
-    ``8``, ``4`` (≈125, 62.5, 31.25 ms at the pooled rate). All strides are ``1`` in convs;
-    temporal resolution changes only via average pooling.
-
-    - **Spatial (how electrodes are processed).**
-    Depthwise convs with ``k=(1,8)`` span all channels and are applied **per temporal branch**,
-    yielding scale-specific channel projections (no cross-branch mixing until concatenation).
-    There is no full 2D mixing kernel; spatial mixing is factorized and lightweight.
-
-    - **Spectral (how frequency information is captured).**
-    No explicit transform; multiple temporal kernels form a *learned filter bank* over
-    ERP-relevant bands. Successive pooling acts as low-pass integration to emphasize sustained
-    post-stimulus components.
-
-    .. rubric:: Additional Mechanisms
-
-    - Every conv/depthwise block includes **BatchNorm**, nonlinearity (paper used grid-searched activation), and **dropout**.
-    - Two Inception stages followed by short convs and pooling keep parameters small (≈15k reported) while preserving multi-scale evidence.
-    - Expected input: epochs of shape ``(B,1,128,8)`` (time x channels as a 2D map) or reshaped from ``(B,8,128)`` with an added singleton feature dimension.
-
-    .. rubric:: Usage and Configuration
-
-    - **Key knobs.** Number of filters per branch; kernel lengths in both Inception modules; depthwise kernel over channels (typically ``n_chans``); pooling lengths/strides; dropout rate; choice of activation.
-    - **Training tips.** Use 0-1000 ms windows at 128 Hz with CAR; tune activation and dropout (they strongly affect performance); early-stop on validation loss when overfitting emerges.
-
-    .. rubric:: Implementation Details
+    The code for the paper and this model is also available at [santamaria2020]_
+    and an adaptation for PyTorch [2]_.
 
     The model is strongly based on the original InceptionNet for an image. The main goal is
     to extract features in parallel with different scales. The authors extracted three scales
@@ -105,9 +33,12 @@ class EEGInceptionERP(EEGModuleMixin, nn.Sequential):
     The winners of BEETL Competition/NeurIps 2021 used parts of the
     model [beetl]_.
 
-    The code for the paper and this model is also available at [santamaria2020]_
-    and an adaptation for PyTorch [2]_.
+    The model is fully described in [santamaria2020]_.
 
+    Notes
+    -----
+    This implementation is not guaranteed to be correct, has not been checked
+    by original authors, only reimplemented from the paper based on [2]_.
 
     Parameters
     ----------

{braindecode-1.2.0.dev175267687 → braindecode-1.2.0.dev175337561}/braindecode/models/eeginception_mi.py

@@ -13,8 +13,6 @@ from braindecode.modules import Ensure4d
 class EEGInceptionMI(EEGModuleMixin, nn.Module):
     """EEG Inception for Motor Imagery, as proposed in Zhang et al. (2021) [1]_
 
-    :bdg-success:`Convolution`
-
     .. figure:: https://content.cld.iop.org/journals/1741-2552/18/4/046014/revision3/jneabed81f1_hr.jpg
         :align: center
         :alt: EEGInceptionMI Architecture