Merge 7f1978b into 3f037dd

.travis.yml

@@ -15,22 +15,32 @@ env:
   - DISTRIB="conda" PYTHON_VERSION="3.5" COVERAGE="true"
 
 install: source build_tools/travis/install.sh
-script: bash build_tools/travis/test.sh
+script:
+  - bash build_tools/travis/test.sh
+  - bash build_tools/travis/make_docs.sh
 
 after_success:
   - source build_tools/travis/success.sh
 #  - chmod +x build_tools/travis/deploy_anaconda.sh
 
 deploy:
 
+  - provider: pages
+    local-dir: doc/_build/html
+    skip-cleanup: true
+    github-token: $GITHUB_TOKEN  # Set in the settings page of your repository, as a secure variable
+    keep-history: true
+    on:
+      branch: master
+
   - provider: pypi
+    skip_cleanup: true
     user: phausamann
     password:
       secure: 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
     on:
       branch: master
       tags: true
-    skip_cleanup: true
 
 #  - provider: script
 #    script: build_tools/travis/deploy_anaconda.sh

README.rst

@@ -1,6 +1,6 @@
 .. -*- mode: rst -*-
 
-|Travis|_ |AppVeyor|_ |CircleCI|_ |Coverage|_ |PyPI|_
+|Travis|_ |AppVeyor|_ |Coverage|_ |PyPI|_
 
 .. |Travis| image:: https://travis-ci.org/phausamann/sklearn-xarray.svg?branch=master
 .. _Travis: https://travis-ci.org/phausamann/sklearn-xarray
@@ -11,9 +11,6 @@
 .. |Coverage| image:: https://coveralls.io/repos/github/phausamann/sklearn-xarray/badge.svg?branch=master
 .. _Coverage: https://coveralls.io/github/phausamann/sklearn-xarray?branch=master
 
-.. |CircleCI| image:: https://circleci.com/gh/phausamann/sklearn-xarray.svg?style=shield&circle-token=:circle-token
-.. _CircleCI: https://circleci.com/gh/phausamann/sklearn-xarray
-
 .. |PyPI| image:: https://badge.fury.io/py/sklearn-xarray.svg
 .. _PyPI: https://badge.fury.io/py/sklearn-xarray
 

build_tools/travis/install.sh

@@ -24,10 +24,9 @@ popd
 
 # Configure the conda environment and put it in the path using the
 # provided versions
-conda create -n testenv --yes python=$PYTHON_VERSION pip nose \
-        --file requirements.txt
-
-# numpy=$NUMPY_VERSION scipy=$SCIPY_VERSION cython=$CYTHON_VERSION
+conda create -n testenv -y -c conda-forge python=$PYTHON_VERSION \
+    nose matplotlib sphinx pillow sphinx-gallery sphinx_rtd_theme numpydoc \
+    --file requirements.txt
 
 source activate testenv
 

build_tools/travis/make_docs.sh

@@ -0,0 +1,10 @@
+mkdir -p doc/modules/generated
+
+cd doc
+set -o pipefail && make html 2>&1 | tee ~/log.txt
+cd ..
+
+cat ~/log.txt && if grep -q "Traceback (most recent call last):" ~/log.txt; \
+    then false; else true; fi
+
+cp .nojekyll doc/_build/html/.nojekyll

build_tools/travis/test.sh

@@ -5,10 +5,13 @@ set -e
 mkdir -p $TEST_DIR
 cp .coveragerc $TEST_DIR/.coveragerc
 
+wd=$(pwd)
 cd $TEST_DIR
 
 if [[ "$COVERAGE" == "true" ]]; then
     nosetests -s --with-coverage --cover-package=$MODULE $MODULE
 else
     nosetests -s $MODULE
 fi
+
+cd $wd

doc/Makefile

@@ -2,7 +2,7 @@
 #
 
 # You can set these variables from the command line.
-SPHINXOPTS    =
+SPHINXOPTS    = -v
 SPHINXBUILD   = sphinx-build
 PAPER         =
 BUILDDIR      = _build

doc/content/whatsnew.rst

@@ -2,6 +2,15 @@ What's New
 ==========
 
 
+Github master branch
+--------------------
+
+Enhancements
+~~~~~~~~~~~~
+
+- Added ``inverse_transform`` to ``preprocessing.Concatenator``.
+
+
 v0.2.0 (April 9, 2018)
 ----------------------
 

examples/plot_activity_recognition.py

@@ -31,7 +31,7 @@
 
 from sklearn.preprocessing import StandardScaler, LabelEncoder
 from sklearn.decomposition import PCA
-from sklearn.linear_model import LogisticRegression
+from sklearn.naive_bayes import GaussianNB
 from sklearn.model_selection import GroupShuffleSplit, GridSearchCV
 from sklearn.pipeline import Pipeline
 
@@ -74,16 +74,16 @@
 # three-dimensional but the standardizer and classifier expect a one-dimensional
 # feature vector, we have to vectorize the samples.
 #
-# Finally, we use PCA and logistic regression to perform the classification.
+# Finally, we use PCA and a naive Bayes classifier for classification.
 
 pl = Pipeline([
     ('splitter', Splitter(
-        groupby=['subject', 'activity'], new_dim='timepoint')),
+        groupby=['subject', 'activity'], new_dim='timepoint', new_len=30)),
     ('sanitizer', Sanitizer()),
     ('featurizer', Featurizer()),
     ('scaler', wrap(StandardScaler)),
     ('pca', wrap(PCA, reshapes='feature')),
-    ('lr', wrap(LogisticRegression, reshapes='feature'))
+    ('cls', wrap(GaussianNB, reshapes='feature'))
 ])
 
 ##############################################################################
@@ -94,20 +94,19 @@
 # training and validation set.
 
 cv = CrossValidatorWrapper(
-    GroupShuffleSplit(n_splits=3, test_size=0.3), groupby=['subject'])
+    GroupShuffleSplit(n_splits=2, test_size=0.5), groupby=['subject'])
 
 ##############################################################################
-# The grid search will try different combinations of segment length and
-# number of PCA components to find the best parameters for this task.
+# The grid search will try different numbers of PCA components to find the best
+# parameters for this task.
 #
 # .. tip::
 #
 #     To use multi-processing, set ``n_jobs=-1``.
 
 gs = GridSearchCV(
     pl, cv=cv, n_jobs=1, verbose=1, param_grid={
-        'splitter__new_len': [30, 60],
-        'pca__n_components': [20, 40]
+        'pca__n_components': [10, 20]
     })
 
 ##############################################################################
@@ -129,5 +128,5 @@
 ##############################################################################
 # .. note::
 #
-#     The performance of this classifier is obviously pretty terrible, it's only
-#     for demonstration purposes.
+#     The performance of this classifier is obviously pretty bad,
+#     it was chosen for execution speed, not accuracy.

sklearn_xarray/common/base.py

@@ -58,7 +58,6 @@ def _update_dims(self, X_in, X_out):
                 raise ValueError(
                     'Inconsistent dimensions returned by estimator')
 
-            # get new dims
             for new_dim, old_dims in self.reshapes.items():
                 for d in old_dims:
                     dims_new.remove(d)
@@ -73,23 +72,71 @@ def _update_dims(self, X_in, X_out):
 
         return dims_new
 
+    def _restore_dims(self, X_in, X_out):
+        """ Restore the dimensions of a reshaped DataArray. """
+
+        # dict syntax
+        if hasattr(self.reshapes, 'items'):
+
+            # check if new dims are dropped by estimator
+            all_old_dims = []
+            for _, old_dims in self.reshapes.items():
+                all_old_dims += old_dims
+
+            if X_in.ndim == X_out.ndim - len(all_old_dims) + len(self.reshapes):
+                drop_new_dims = False
+            elif X_in.ndim == X_out.ndim - len(all_old_dims):
+                drop_new_dims = True
+            else:
+                raise ValueError(
+                    'Inconsistent dimensions returned by estimator')
+
+            # get new dims
+            dims_new = list(X_in.dims)
+            dims_old = []
+            for d in dims_new:
+                if d in self.reshapes:
+                    dims_old += self.reshapes[d]
+                else:
+                    dims_old.append(d)
+
+            if drop_new_dims:
+                # TODO: figure out where to insert the dropped dims
+                for d in all_old_dims:
+                    if d not in dims_old:
+                        dims_old.append(d)
+
+        # string syntax
+        else:
+            dims_old = list(X_in.dims)
+            # check if dim is dropped by estimator
+            if X_out.ndim < X_in.ndim:
+                # TODO: figure out where to insert the dropped dim
+                dims_old.append(self.reshapes)
+
+        return dims_old
+
     def _update_coords(self, X):
         """ Update the coordinates of a reshaped DataArray. """
 
         coords_new = dict()
 
         # dict syntax
         if hasattr(self.reshapes, 'items'):
-            # drop all coords along the reshaped dimensions
+
+            all_old_dims = []
             for _, old_dims in self.reshapes.items():
-                for c in X.coords:
-                    old_dims_in_c = [x for x in X[c].dims if x in old_dims]
-                    if any(old_dims_in_c) and c not in old_dims:
-                        c_t = X[c].isel(**{d: 0 for d in old_dims_in_c})
-                        new_dims = [d for d in X[c].dims if d not in old_dims]
-                        coords_new[c] = (new_dims, c_t.drop(old_dims_in_c))
-                    elif c not in old_dims:
-                        coords_new[c] = X[c]
+                all_old_dims += old_dims
+
+            # drop all coords along the reshaped dimensions
+            for c in X.coords:
+                old_dims_in_c = [x for x in X[c].dims if x in all_old_dims]
+                if any(old_dims_in_c) and c not in all_old_dims:
+                    c_t = X[c].isel(**{d: 0 for d in old_dims_in_c})
+                    new_dims = [d for d in X[c].dims if d not in all_old_dims]
+                    coords_new[c] = (new_dims, c_t.drop(old_dims_in_c))
+                elif c not in all_old_dims:
+                    coords_new[c] = X[c]
 
         # string syntax
         else:
@@ -117,8 +164,11 @@ def _call_array_method(self, estimator, method, X):
         else:
             y = getattr(estimator, method)(X.data)
 
-        # update coords and dims
-        dims_new = self._update_dims(X, y)
+        # update dims
+        if method == 'inverse_transform':
+            dims_new = self._restore_dims(X, y)
+        else:
+            dims_new = self._update_dims(X, y)
 
         return y, dims_new