Merge pull request #74 from wwu-mmll/develop

Develop

Author: jernsting

.github/workflows/python-test_and_deploy.yml

@@ -12,38 +12,6 @@ on:
     - cron: '0 10 5 * *'
 
 jobs:
-  pytest:
-    name: Run PHOTONAI tests ${{matrix.python-version}}
-    runs-on: ubuntu-22.04
-    
-    services:
-      mongodb:
-        image: mongo:latest
-        ports:
-          - 27017:27017
-
-    strategy:
-      matrix:
-        python-version: ['3.8', '3.9']
-
-    steps:
-    - uses: actions/checkout@v3
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v4
-      with:
-        python-version: ${{ matrix.python-version }}
-#    - name: Install os dependencies
-#      run: | 
-#        sudo apt-get update
-#        sudo apt-get -y install gfortran swig
-    - name: Install dependencies
-      run: |
-        pip install wheel flake8 pbr
-        python setup.py egg_info
-        pip install tensorflow pytest pytest-cov coveralls -r photonai.egg-info/requires.txt -r photonai/optimization/smac/requirements.txt -r photonai/optimization/nevergrad/requirements.txt
-    - name: Test with pytest
-      run: |
-        PYTHONPATH=./ pytest ./test --cov=./photonai --full-trace
   coverage:
     name: Run PHOTONAI tests and publish test coverage
     runs-on: ubuntu-22.04
@@ -56,14 +24,10 @@ jobs:
 
     steps:
     - uses: actions/checkout@v3
-    - name: Set up Python ${{ matrix.python-version }}
+    - name: Set up Python 3.9
       uses: actions/setup-python@v4
       with:
-        python-version: ${{ matrix.python-version }}
-    #    - name: Install os dependencies
-    #      run: |
-    #        sudo apt-get update
-    #        sudo apt-get -y install gfortran swig
+        python-version: 3.9
     - name: Install dependencies
       run: |
         pip install wheel flake8 pbr
@@ -74,23 +38,21 @@ jobs:
         PYTHONPATH=./ pytest ./test --cov=./photonai --full-trace
     - name: Coveralls
       run: coveralls
-      if: ${{ matrix.python-version }} == 3.10
       env:
         COVERALLS_REPO_TOKEN: ${{ secrets.COVERALLS_REPO_TOKEN }}
         GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   deploy:
     name: Build and publish to TestPyPI
     runs-on: ubuntu-22.04
-    needs: pytest
     if: github.ref == 'refs/heads/main'
     steps:
     - uses: actions/checkout@v3
       with:
         fetch-depth: 0
-    - name: Set up Python 3.10
+    - name: Set up Python 3.9
       uses: actions/setup-python@v4
       with:
-        python-version: 3.10.8
+        python-version: 3.9
     - name: Install pypa/build
       run: pip install build pbr wheel
     - name: Build a binary wheel and a source tarball

.gitignore

@@ -10,6 +10,9 @@ site/
 **/cache
 **/dask-worker-space
 
+**/classification
+**/regression
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

documentation/docs/api/base/classification_pipe.md

@@ -0,0 +1,5 @@
+# Documentation for `ClassificationPipe`
+::: photonai.base.ClassificationPipe
+    selection:
+      members:
+        - __init__ 

documentation/docs/api/base/classifier_switch.md

@@ -0,0 +1,5 @@
+# Documentation for `ClassifierSwitch`
+::: photonai.base.ClassifierSwitch
+    selection:
+      members:
+        - __init__ 

documentation/docs/api/base/regression_pipe.md

@@ -0,0 +1,5 @@
+# Documentation for `RegressionPipe`
+::: photonai.base.RegressionPipe
+    selection:
+      members:
+        - __init__ 

documentation/docs/api/base/regressor_switch.md

@@ -0,0 +1,5 @@
+# Documentation for `RegressorSwitch`
+::: photonai.base.RegressorSwitch
+    selection:
+      members:
+        - __init__ 

documentation/docs/examples/classification.md documentation/docs/examples/default_classification.md

@@ -6,7 +6,16 @@
     </p>
 </div>
 
+<h3>Classification with default pipeline</h3>
+
 ``` python
 {% include "examples/basic/classification.py" %} 
 
+```
+
+<h3>Classification with custom pipeline</h3>
+
+``` python
+{% include "examples/basic/classification_custom.py" %} 
+
 ```

documentation/docs/getting_started/classification.md

@@ -6,7 +6,15 @@
     </p>
 </div>
 
+<h3>Regression with default pipeline</h3>
 ``` python
 {% include "examples/basic/regression.py" %} 
 
+```
+
+<h3>Regression with custom pipeline</h3>
+
+``` python
+{% include "examples/basic/regression_custom.py" %} 
+
 ```

documentation/docs/getting_started/regression.md

@@ -1,31 +1,55 @@
 <h1>Getting Started</h1>
-<h2>1. Installation</h2>
+<h2>Installation</h2>
 <p class="small-p">You only need two things: Python 3 and your favourite Python IDE to get started. Then simply install via pip.</p>
 
 ```python
 pip install photonai
 ```
 
-<h2>2. Setup New Analysis</h2>
-Start by importing some utilities and creating a new Hyperpipe instance, naming the analysis and specifying where to save all outputs. 
+<h2> Setup Default Analysis </h2>
+<h3>Regression with default pipeline</h3>
 
 ```python
-from sklearn.model_selection import ShuffleSplit, KFold
+from sklearn.datasets import load_diabetes
+from photonai import RegressionPipe
+
+my_pipe = RegressionPipe('diabetes')
+X, y = load_diabetes(return_X_y=True)
+my_pipe.fit(X, y)
+```
+
+<h3>Classification with modified default pipeline</h2>
+``` python
+from photonai import ClassificationPipe
 from sklearn.datasets import load_breast_cancer
-from photonai.base import Hyperpipe, PipelineElement, Switch
-from photonai.optimization import IntegerRange, FloatRange
+from sklearn.model_selection import ShuffleSplit
 
-pipe = Hyperpipe('basic_pipe', project_folder='./')
+X, y = load_breast_cancer(return_X_y=True)
+my_pipe = ClassificationPipe(name='breast_cancer_analysis',
+                             inner_cv=ShuffleSplit(n_splits=2),
+                             scaling=True,
+                             imputation=False,
+                             imputation_nan_value=None,
+                             feature_selection=False,
+                             dim_reduction=True,
+                             n_pca_components=10)
+my_pipe.fit(X, y)
 ```
 
 
-<h2>3. Define training, optimization and testing parameters</h2>
-Select parameters to customize the training, hyperparameter optimization and testing procedure.
+<h2>Or Setup New Custom Analysis</h2>
 
+Start by importing some utilities. and creating a new Hyperpipe instance, naming the analysis and specifying where to save all outputs.
+Select parameters to customize the training, hyperparameter optimization and testing procedure.
 Particularly, you can choose the hyperparameter optimization strategy, set parameters, choose performance metrics
 and choose the performance metric to minimize or maximize, respectively.
 
-```python 
+```python
+from sklearn.model_selection import ShuffleSplit, KFold
+from sklearn.datasets import load_breast_cancer
+from photonai.base import Hyperpipe, PipelineElement, Switch
+from photonai.optimization import IntegerRange, FloatRange
+
 pipe = Hyperpipe('basic_pipe', project_folder='./',
 
                   # choose hyperparameter optimization strategy
@@ -43,7 +67,7 @@ pipe = Hyperpipe('basic_pipe', project_folder='./',
                   inner_cv=KFold(n_splits=10))
 ``` 
 
-<h2>4. Build custom pipeline</h2>
+<h3>Add pipeline elements to your liking.</h3>
 Select and arrange normalization, dimensionality reduction, feature selection, data augmentation,
 over- or undersampling algorithms in simple or parallel data streams. You can integrate
 custom algorithms or choose from our wide range of pre-registered algorithms from established toolboxes.
@@ -69,7 +93,7 @@ or_element += PipelineElement('SVC',
 pipe += or_element
 ```
 
-<h2>5. Load Data and Train</h2> 
+<h3>Finally, load data and start training!</h3> 
 Load your data and start the (nested-) cross-validated hyperparameter optimization, training and evaluation procedure.
 You will see an extensive output to monitor the hyperparameter optimization progress, see the results and track the
 best performances so far.

documentation/docs/index.md

@@ -18,6 +18,7 @@
 my_pipe += PipelineElement('Ridge', alpha=1e-2)
 my_pipe.fit(X_train, y_train)
 
+# get model-agnostic feature importances
 r = my_pipe.get_permutation_feature_importances(n_repeats=50, random_state=0)
 
 for i in r["mean"].argsort()[::-1]:
@@ -26,7 +27,9 @@
               f"{r['mean'][i]:.3f}"
               f" +/- {r['std'][i]:.3f}")
 
+# if your have a model that returns feature importance scores you can compare those to permutation feature importances
+feature_importances_df = my_pipe.get_model_and_permutation_importances(n_repeats=50, random_state=0)
 
 # get permutation importances posthoc
 # reloaded_hyperpipe = Hyperpipe.reload_hyperpipe("full_path/to/results_folder/", X_train, y_train)
-# post_hoc_perm_importances = Hyperpipe.get_permutation_feature_importances(n_repeats=5, random_state=0)
+# post_hoc_perm_importances = reloaded_hyperpipe.get_permutation_feature_importances(n_repeats=5, random_state=0)

examples/advanced/feature_importance.py

@@ -1,28 +1,15 @@
+from photonai import ClassificationPipe
 from sklearn.datasets import load_breast_cancer
-from sklearn.model_selection import KFold
+from sklearn.model_selection import ShuffleSplit
 
-from photonai import Hyperpipe, PipelineElement, FloatRange, Categorical, IntegerRange
-
-my_pipe = Hyperpipe('basic_svm_pipe',
-                    inner_cv=KFold(n_splits=5),
-                    outer_cv=KFold(n_splits=3),
-                    optimizer='sk_opt',
-                    optimizer_params={'n_configurations': 15},
-                    metrics=['accuracy', 'precision', 'recall', 'balanced_accuracy'],
-                    best_config_metric='accuracy',
-                    project_folder='./tmp')
-
-my_pipe.add(PipelineElement('StandardScaler'))
-
-my_pipe += PipelineElement('PCA',
-                           hyperparameters={'n_components': IntegerRange(10, 30)},
-                           n_components=10,
-                           test_disabled=True)
-
-my_pipe += PipelineElement('SVC',
-                           hyperparameters={'kernel': Categorical(['rbf', 'linear']),
-                                            'C': FloatRange(1, 6)},
-                           gamma='scale')
 
 X, y = load_breast_cancer(return_X_y=True)
+my_pipe = ClassificationPipe(name='breast_cancer_analysis',
+                             inner_cv=ShuffleSplit(n_splits=2),
+                             scaling=True,
+                             imputation=False,
+                             imputation_nan_value=None,
+                             feature_selection=False,
+                             dim_reduction=True,
+                             n_pca_components=10)
 my_pipe.fit(X, y)

examples/basic/classification.py

@@ -0,0 +1,28 @@
+from sklearn.model_selection import KFold
+from sklearn.datasets import load_breast_cancer
+from photonai import Hyperpipe, PipelineElement, FloatRange, Categorical, IntegerRange
+
+my_pipe = Hyperpipe('basic_svm_pipe',
+                    inner_cv=KFold(n_splits=5),
+                    outer_cv=KFold(n_splits=3),
+                    optimizer='sk_opt',
+                    optimizer_params={'n_configurations': 15},
+                    metrics=['accuracy', 'precision', 'recall', 'balanced_accuracy'],
+                    best_config_metric='accuracy',
+                    project_folder='./tmp')
+
+my_pipe.add(PipelineElement('StandardScaler'))
+
+my_pipe += PipelineElement('PCA',
+                           hyperparameters={'n_components': IntegerRange(10, 30)},
+                           n_components=10,
+                           test_disabled=True)
+
+my_pipe += PipelineElement('SVC',
+                           hyperparameters={'kernel': Categorical(['rbf', 'linear']),
+                                            'C': FloatRange(1, 6)},
+                           gamma='scale')
+
+
+X, y = load_breast_cancer(return_X_y=True)
+my_pipe.fit(X, y)

examples/basic/classification_custom.py

@@ -0,0 +1,26 @@
+from sklearn.model_selection import KFold
+from sklearn.datasets import load_breast_cancer
+from photonai import Hyperpipe, PipelineElement
+
+my_pipe = Hyperpipe('basic_svm_pipe',
+                    inner_cv=KFold(n_splits=5),
+                    outer_cv=KFold(n_splits=3),
+                    metrics=['balanced_accuracy',
+                             'precision',
+                             'recall',
+                             'accuracy'],
+                    best_config_metric='balanced_accuracy',
+                    project_folder='./tmp')
+
+my_pipe += PipelineElement('StandardScaler')
+
+my_pipe += PipelineElement('PCA',
+                           n_components=10,
+                           test_disabled=True)
+
+my_pipe += PipelineElement('LogisticRegression')
+
+
+X, y = load_breast_cancer(return_X_y=True)
+my_pipe.fit(X, y)
+

examples/basic/no_hyperparameters.py

@@ -1,29 +1,15 @@
 from sklearn.datasets import load_diabetes
-from sklearn.model_selection import KFold
-
-from photonai import Hyperpipe, PipelineElement, IntegerRange, FloatRange
-
-
-my_pipe = Hyperpipe('basic_regression_pipe',
-                    optimizer='random_search',
-                    optimizer_params={'n_configurations': 25},
-                    metrics=['mean_squared_error', 'mean_absolute_error', 'explained_variance'],
-                    best_config_metric='mean_squared_error',
-                    outer_cv=KFold(n_splits=3, shuffle=True),
-                    inner_cv=KFold(n_splits=3, shuffle=True),
-                    verbosity=1,
-                    project_folder='./tmp/')
-
-my_pipe += PipelineElement('SimpleImputer')
-my_pipe += PipelineElement('StandardScaler')
-
-my_pipe += PipelineElement('LassoFeatureSelection',
-                           hyperparameters={'percentile': [0.1, 0.2, 0.3],
-                                            'alpha': FloatRange(0.5, 5)})
-
-my_pipe += PipelineElement('RandomForestRegressor',
-                           hyperparameters={'n_estimators': IntegerRange(10, 50)})
-
+from photonai import RegressionPipe
+
+my_pipe = RegressionPipe('diabetes',
+                         add_default_pipeline_elements=True,
+                         scaling=True,
+                         imputation=False,
+                         imputation_nan_value=None,
+                         feature_selection=False,
+                         dim_reduction=True,
+                         n_pca_components=10,
+                         add_estimator=True)
 # load data and train
 X, y = load_diabetes(return_X_y=True)
 my_pipe.fit(X, y)