Adding Braindecode pipeline

.coveragerc

@@ -4,6 +4,7 @@ source = moabb
 omit =
     */docs/*
     */pipelines/*
+    */tests/*
 
 [report]
 exclude_lines =

docs/source/whats_new.rst

@@ -27,6 +27,7 @@ Enhancements
 - Add a augmentation model to the pipeline (:gh:`326` by `Igor Carrara`_)
 - Add BrainDecode example(:gh:`340` by `Igor Carrara`_ and `Bruno Aristimunha`_)
 - Add Google Analytics to the documentation (:gh:`335` by `Bruno Aristimunha`_)
+- Add suport to Braindecode classifier (:gh:`328` by `Bruno Aristimunha`_)
 - Add CodeCarbon to track emission CO₂ (:gh:`350` by `Igor Carrara`_, `Bruno Aristimunha`_ and `Sylvain Chevallier`_)
 - Add CodeCarbon example (:gh:`356` by `Igor Carrara`_ and `Bruno Aristimunha`_)
 

examples/pipelines_braindecode/braindecode_EEGInception.py

@@ -22,10 +22,10 @@
 PATIENCE = 3
 
 # Create the dataset
-create_dataset = BraindecodeDatasetLoader()
+create_dataset = BraindecodeDatasetLoader(drop_last_window=False)
 
 # Set EEG Inception model
-model = EEGInception(in_channels=1, n_classes=2, input_window_samples=100)
+model = EEGInception(in_channels=1, n_classes=2)
 
 # Define a Skorch classifier
 clf = EEGClassifier(

examples/pipelines_braindecode/braindecode_EEGNetv4.py

@@ -22,7 +22,7 @@
 PATIENCE = 3
 
 # Create the dataset
-create_dataset = BraindecodeDatasetLoader()
+create_dataset = BraindecodeDatasetLoader(drop_last_window=False)
 
 # Set EEGNetv4 model
 model = EEGNetv4(in_chans=1, n_classes=2, input_window_samples=100)

examples/pipelines_braindecode/braindecode_ShallowFBCSPNet.py

@@ -22,7 +22,7 @@
 PATIENCE = 3
 
 # Create the dataset
-create_dataset = BraindecodeDatasetLoader()
+create_dataset = BraindecodeDatasetLoader(drop_last_window=False)
 
 # Set Shallow Filter Bank CSP Net model
 model = ShallowFBCSPNet(

moabb/benchmark.py

@@ -32,7 +32,7 @@
 log = logging.getLogger(__name__)
 
 
-def benchmark(
+def benchmark(  # noqa: C901
     pipelines="./pipelines/",
     evaluations=None,
     paradigms=None,
@@ -154,27 +154,49 @@ def benchmark(
                 f"Datasets considered for {paradigm} paradigm {[dt.code for dt in d]}"
             )
 
-            if "braindecode" in list(prdgms[paradigm].keys())[0]:
-                return_epochs = True
-            else:
-                return_epochs = False
-
-            context = eval_type[evaluation](
-                paradigm=p,
-                datasets=d,
-                random_state=42,
-                hdf5_path=results,
-                n_jobs=n_jobs,
-                overwrite=overwrite,
-                return_epochs=return_epochs,
-            )
-            paradigm_results = context.process(
-                pipelines=prdgms[paradigm], param_grid=param_grid
-            )
-            paradigm_results["paradigm"] = f"{paradigm}"
-            paradigm_results["evaluation"] = f"{evaluation}"
-            eval_results[f"{paradigm}"] = paradigm_results
-            df_eval.append(paradigm_results)
+            ppl_with_epochs, ppl_with_array = {}, {}
+            for pn, pv in prdgms[paradigm].items():
+                if "braindecode" in pn:
+                    ppl_with_epochs[pn] = pv
+                else:
+                    ppl_with_array[pn] = pv
+
+            if len(ppl_with_epochs) > 0:
+                # Braindecode pipelines require return_epochs=True
+                context = eval_type[evaluation](
+                    paradigm=p,
+                    datasets=d,
+                    random_state=42,
+                    hdf5_path=results,
+                    n_jobs=1,
+                    overwrite=overwrite,
+                    return_epochs=True,
+                )
+                paradigm_results = context.process(
+                    pipelines=ppl_with_epochs, param_grid=param_grid
+                )
+                paradigm_results["paradigm"] = f"{paradigm}"
+                paradigm_results["evaluation"] = f"{evaluation}"
+                eval_results[f"{paradigm}"] = paradigm_results
+                df_eval.append(paradigm_results)
+
+            # Other pipelines, that use numpy arrays
+            if len(ppl_with_array) > 0:
+                context = eval_type[evaluation](
+                    paradigm=p,
+                    datasets=d,
+                    random_state=42,
+                    hdf5_path=results,
+                    n_jobs=n_jobs,
+                    overwrite=overwrite,
+                )
+                paradigm_results = context.process(
+                    pipelines=ppl_with_array, param_grid=param_grid
+                )
+                paradigm_results["paradigm"] = f"{paradigm}"
+                paradigm_results["evaluation"] = f"{evaluation}"
+                eval_results[f"{paradigm}"] = paradigm_results
+                df_eval.append(paradigm_results)
 
         # Combining FilterBank and direct paradigms
         eval_results = _combine_paradigms(eval_results)

moabb/pipelines/__init__.py

@@ -3,7 +3,9 @@
 Pipelines are typically a chain of sklearn compatible transformers and end
 with a sklearn compatible estimator.
 """
+
 # flake8: noqa
+
 from .classification import SSVEP_CCA, SSVEP_TRCA
 from .features import FM, AugmentedDataset, ExtendedSSVEPSignal, LogVariance
 from .utils import FilterBank, create_pipeline_from_config
@@ -20,7 +22,7 @@
     )
     from .utils_deep_model import EEGNet, TCN_block
 except ModuleNotFoundError as err:
-    print("Tensorflow not install, you could not use deep learning pipelines")
+    print("Tensorflow not install, you could not use those pipelines")
 
 try:
     from .utils_pytorch import (

moabb/pipelines/utils_pytorch.py

@@ -2,32 +2,58 @@
 from functools import partial
 from inspect import getmembers, isclass, isroutine
 
+import mne
 from braindecode.datasets import BaseConcatDataset, create_from_X_y
 from numpy import unique
 from sklearn.base import BaseEstimator, TransformerMixin
 from skorch.callbacks import Callback
 from torch.nn import Module
 
 
+# check if the data format is numpy or mne epoch
+def _check_data_format(X):
+    """
+    Check if the data format is compatible with braindecode.
+    Expect values in the format of MNE objects.
+    Parameters
+    ----------
+    X: BaseConcatDataset
+
+    Returns
+    -------
+
+    """
+    if not isinstance(X, mne.EpochsArray):
+        raise ValueError(
+            "The data format is not supported. "
+            "Please use the option return_epochs=True"
+            "inside the Evaluations module."
+        )
+
+
 class BraindecodeDatasetLoader(BaseEstimator, TransformerMixin):
     """
     Class to Load the data from MOABB in a format compatible with braindecode
     """
 
-    def __init__(self, kw_args=None):
+    def __init__(self, drop_last_window=False, kw_args=None):
+        self.drop_last_window = drop_last_window
         self.kw_args = kw_args
 
     def fit(self, X, y=None):
+        _check_data_format(X)
         self.y = y
         return self
 
     def transform(self, X, y=None):
+        _check_data_format(X)
         dataset = create_from_X_y(
-            X.get_data(),
+            X=X.get_data(),
             y=self.y,
             window_size_samples=X.get_data().shape[2],
             window_stride_samples=X.get_data().shape[2],
-            drop_last_window=False,
+            drop_last_window=self.drop_last_window,
+            ch_names=X.info["ch_names"],
             sfreq=X.info["sfreq"],
         )
 

moabb/tests/util_braindecode.py

@@ -112,6 +112,12 @@ def test_type_create_from_X_y_vs_transfomer(self, data):
         assert isinstance(dataset_trans, BaseConcatDataset)
         assert type(dataset_trans) == type(dataset)
 
+    def test_wrong_input(self):
+        """Test that an invalid input raises a ValueError"""
+        transformer = BraindecodeDatasetLoader()
+        with pytest.raises(ValueError):
+            transformer.fit_transform(np.random.normal(size=(2, 1, 10)), y=np.array([0]))
+
 
 if __name__ == "__main__":
     unittest.main()

moabb/utils.py

@@ -102,11 +102,11 @@ def setup_seed(seed: int) -> None:
 
 
 def set_log_level(level="INFO"):
-    """Set lot level.
+    """Set log level
 
     Set the general log level.
     Use one of the levels supported by python logging, i.e.:
-        DEBUG, INFO, WARNING, ERROR, CRITICAL
+    DEBUG, INFO, WARNING, ERROR, CRITICAL
     """
     VALID_LEVELS = ["DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"]
     level = level.upper()

pyproject.toml

@@ -56,6 +56,7 @@ pre-commit = "^2.21.0"
 m2r2 = "^0.3.3"
 tdlda = {git = "https://github.com/jsosulski/tdlda.git", rev = "0.1.0"}
 
+
 [build-system]
 requires = ["poetry-core>=1.0.0"]
 build-backend = "poetry.core.masonry.api"