Added an example that uses datasets from MOABB and the RG+QuantumSVM …

…pipeline (#39)

* Create a new workflow.

* This script uses the new version of MOABB that now includes several Brain Invaders datasets.
It classifies using a Quantum Classifier and Riemannian geometry.

* Small improvement.

* Fixed a bug which forced classical SVM instead of Quantum.

* force classical SVM

* Improved code that compares 3 pipelines.

* Now it prints the result in the console.

* Comments are improved.

* comment

* Added moabb 0.4.6 as a requirement.

* Delete python-package.yml

* Some updates to comply with flake 8.

* update

* update

* description update

* Added version of pure SVM on the raw data.

* Improvements.
The number of pipelines has been reduced to two. This way it will be more simple to compare results.

* add example to index.rst
add `print(__doc__)` to example

* flake8

* ignore python version (3.6 not supported in moabb)

* drop support for 3.6 and add support for 3.9

* drop support for 3.9

* add comments to Vectorizer

* try installing moabb from the source

Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

* re-test python 3.6

* Revert "re-test python 3.6"

This reverts commit f6b20b3.

* Revert "try installing moabb from the source"

This reverts commit f5ecbac.

* fix comment

* add moabb to `docs` dependency.

* re-enable python 3.6

* Trigger pipeline

* comments updated.

* fixing flake8 ...

* remove support for 3.6

* Improved description and applied one suggestion.

* Improved comments.

* flake update

* Fixed typo.

* add_javascript deprecated in Sphinx

* Update classify_P300_bi.py

Change the way `q_account_token` changed (updated in last PR)

Co-authored-by: gcattan <gregoire.cattan@ibm.com>
Co-authored-by: gcattan <gcattan@hotmail.fr>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

.github/workflows/testing.yml

@@ -16,7 +16,7 @@ jobs:
  strategy:
  fail-fast: false
  matrix:
- python-version: [3.6, 3.7, 3.8]
+ python-version: [3.7, 3.8]
  os: [ubuntu-latest, macOS-latest, windows-latest]
 
  steps:
@@ -37,3 +37,4 @@ jobs:
  - name: Test with pytest
  run: |
  pytest
+

doc/conf.py

@@ -331,5 +331,5 @@
 
 
 def setup(app):
- app.add_javascript('copybutton.js')
+ app.add_js_file('copybutton.js')
  app.add_stylesheet('style.css')

doc/index.rst

@@ -32,6 +32,11 @@ pyRiemann-qiskit: Qiskit wrapper for pyRiemann
  <img src="_images/sphx_glr_plot_single_thumb.png">
  </div>
  </a>
+ <a href="auto_examples/ERP/classify_P300_bi.html">
+ <div class="col-md-3 thumbnail">
+ <img src="_images/sphx_glr_plot_single_thumb.png">
+ </div>
+ </a>
  <a href="auto_examples/ERP/compare_dim_red.html">
  <div class="col-md-3 thumbnail">
  <img src="_images/sphx_glr_plot_single_thumb.png">

doc/requirements.txt

@@ -10,4 +10,5 @@ pandas
 cvxpy==1.1.12
 qiskit==0.20.0
 scipy==1.7.3
+moabb>=0.4.6
 git+https://github.com/pyRiemann/pyRiemann#egg=pyriemann

examples/ERP/classify_P300_bi.py

@@ -0,0 +1,148 @@
+"""
+====================================================================
+Classification of P300 datasets from MOABB
+====================================================================
+
+It demonstrates the QuantumClassifierWithDefaultRiemannianPipeline(). This
+pipeline uses Riemannian Geometry, Tangent Space and a quantum SVM
+classifier. MOABB is used to access many EEG datasets and also for the
+evaluation and comparison with other classifiers.
+
+In QuantumClassifierWithDefaultRiemannianPipeline():
+If parameter "shots" is None then a classical SVM is used similar to the one
+in scikit learn.
+If "shots" is not None and IBM Qunatum token is provided with "q_account_token"
+then a real Quantum computer will be used.
+You also need to adjust the "n_components" in the PCA procedure to the number
+of qubits supported by the real quantum computer you are going to use.
+A list of real quantum computers is available in your IBM quantum account.
+
+"""
+# Author: Anton Andreev
+# Modified from plot_classify_EEG_tangentspace.py of pyRiemann
+# License: BSD (3-clause)
+
+from pyriemann.estimation import XdawnCovariances
+from pyriemann.tangentspace import TangentSpace
+from sklearn.pipeline import make_pipeline
+from matplotlib import pyplot as plt
+import warnings
+import seaborn as sns
+from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
+from moabb import set_log_level
+from moabb.datasets import bi2012
+from moabb.evaluations import WithinSessionEvaluation
+from moabb.paradigms import P300
+from pyriemann_qiskit.classification import \
+ QuantumClassifierWithDefaultRiemannianPipeline
+from sklearn.decomposition import PCA
+
+print(__doc__)
+
+##############################################################################
+# getting rid of the warnings about the future
+warnings.simplefilter(action="ignore", category=FutureWarning)
+warnings.simplefilter(action="ignore", category=RuntimeWarning)
+
+warnings.filterwarnings("ignore")
+
+set_log_level("info")
+
+##############################################################################
+# Create Pipelines
+# ----------------
+#
+# Pipelines must be a dict of sklearn pipeline transformer.
+
+##############################################################################
+# We have to do this because the classes are called 'Target' and 'NonTarget'
+# but the evaluation function uses a LabelEncoder, transforming them
+# to 0 and 1
+labels_dict = {"Target": 1, "NonTarget": 0}
+
+paradigm = P300(resample=128)
+
+datasets = [bi2012()] # MOABB provides several other P300 datasets
+
+# reduce the number of subjects, the Quantum pipeline takes a lot of time
+# if executed on the entire dataset
+n_subjects = 5
+for dataset in datasets:
+ dataset.subject_list = dataset.subject_list[0:n_subjects]
+
+overwrite = True # set to True if we want to overwrite cached results
+
+pipelines = {}
+
+# A Riemannian Quantum pipeline provided by pyRiemann-qiskit
+# You can choose between classical SVM and Quantum SVM.
+pipelines["RG+QuantumSVM"] = QuantumClassifierWithDefaultRiemannianPipeline(
+ shots=None, # 'None' forces classic SVM
+ nfilter=2, # default 2
+ # default n_components=10, a higher value renders better performance with
+ # the non-qunatum SVM version used in qiskit
+ # On a real Quantum computer (n_components = qubits)
+ dim_red=PCA(n_components=5),
+ # params={'q_account_token': '<IBM Quantum TOKEN>'}
+ )
+
+# Here we provide a pipeline for comparison:
+
+# This is a standard pipeline similar to
+# QuantumClassifierWithDefaultRiemannianPipeline, but with LDA classifier
+# instead.
+pipelines["RG+LDA"] = make_pipeline(
+ # applies XDawn and calculates the covariance matrix, output it matrices
+ XdawnCovariances(
+ nfilter=2,
+ classes=[labels_dict["Target"]],
+ estimator="lwf",
+ xdawn_estimator="scm"
+ ),
+ TangentSpace(),
+ PCA(n_components=10),
+ LDA(solver="lsqr", shrinkage="auto"), # you can use other classifiers
+)
+
+print("Total pipelines to evaluate: ", len(pipelines))
+
+evaluation = WithinSessionEvaluation(
+ paradigm=paradigm,
+ datasets=datasets,
+ suffix="examples",
+ overwrite=overwrite
+)
+
+results = evaluation.process(pipelines)
+
+print("Averaging the session performance:")
+print(results.groupby('pipeline').mean('score')[['score', 'time']])
+
+##############################################################################
+# Plot Results
+# ----------------
+#
+# Here we plot the results to compare the two pipelines
+
+fig, ax = plt.subplots(facecolor="white", figsize=[8, 4])
+
+sns.stripplot(
+ data=results,
+ y="score",
+ x="pipeline",
+ ax=ax,
+ jitter=True,
+ alpha=0.5,
+ zorder=1,
+ palette="Set1",
+)
+sns.pointplot(data=results,
+ y="score",
+ x="pipeline",
+ ax=ax, zorder=1,
+ palette="Set1")
+
+ax.set_ylabel("ROC AUC")
+ax.set_ylim(0.3, 1)
+
+plt.show()

pyriemann_qiskit/utils/filtering.py

@@ -1,4 +1,5 @@
-from sklearn.base import TransformerMixin
+from sklearn.base import BaseEstimator, TransformerMixin
+import numpy as np
 
 
 class NoDimRed(TransformerMixin):
@@ -112,3 +113,53 @@ def transform(self, X):
  """
  offset = 0 if self.is_even else 1
  return X[:, offset::2]
+
+
+class Vectorizer(BaseEstimator, TransformerMixin):
+ """This is an auxiliary transformer that allows one to vectorize data
+ structures in a pipeline For instance, in the case of an X with
+ dimensions (n_samples, n_features, n_channels),
+ one might be interested in a new data structure with dimensions
+ (n_samples, n_features x n_channels)
+
+ Notes
+ -----
+ .. versionadded:: 0.0.1
+
+ """
+ def __init__(self):
+ pass
+
+ def fit(self, X, y):
+ """Fit the training data.
+
+ Parameters
+ ----------
+ X : ndarray, shape (n_samples, n_features)
+ Training vector, where `n_samples` is the number of samples and
+ `n_features` is the number of features.
+ y : ndarray, shape (n_samples,) (default: None)
+ Target vector relative to X.
+ In practice, never used.
+
+ Returns
+ -------
+ self : Vectorizer instance
+ The Vectorizer instance.
+ """
+ return self
+
+ def transform(self, X):
+ """Vectorize matrices.
+
+ Parameters
+ ----------
+ X : ndarray, shape (n_samples, n_features, n_channels)
+ ndarray of feature vectors.
+
+ Returns
+ -------
+ X : ndarray, shape (n_samples, n_features x n_channels)
+ The vectorized matrices.
+ """
+ return np.reshape(X, (X.shape[0], -1))

setup.py

@@ -40,7 +40,7 @@
  'docplex',
  'tqdm'
  ],
- extras_require={'docs': ['sphinx-gallery', 'sphinx-bootstrap_theme', 'numpydoc', 'mne', 'seaborn'],
+ extras_require={'docs': ['sphinx-gallery', 'sphinx-bootstrap_theme', 'numpydoc', 'mne', 'seaborn', 'moabb>=0.4.6'],
  'tests': ['pytest', 'seaborn', 'flake8', 'mne', 'pooch', 'tqdm']},
  zip_safe=False,
 )