Merge branch 'develop'

docs/source/releases.rst

@@ -1,6 +1,21 @@
 Release Notes
 =============
 
+Version 19.11.0
+---------------
+Features:
+ - `gama.__version__` can now be used to retrieve gama's version.
+ - `fit_arff`, `score_arff` and `predict_arff` now accept a `target_column` parameter to specify the target.
+   If left unset, the last column of the ARFF file is assumed to be the target column.
+
+Bugfixes:
+ - fit(x, y) may now be called with y as (N,1) array.
+ - ensemble post-processing is now compatible with non-zero indexed class labels
+
+Maintenance:
+ - `__version__.py` is now the only place with hard-coded version.
+
+
 Version 19.08.0
 ---------------
 - Prototype dash app for visualizing GAMA logs.

gama/__init__.py

@@ -1,5 +1,6 @@
 from .GamaClassifier import GamaClassifier
 from .GamaRegressor import GamaRegressor
+from .__version__ import __version__
 
 name = "gama"
 

gama/__version__.py

@@ -0,0 +1,2 @@
+# We employ YY.0M.micro scheme in 2019. In 2020 we move to YY.minor.micro.
+__version__ = '19.11.0'

gama/data.py

@@ -1,5 +1,5 @@
 """ This module contains functions for loading data. """
-from typing import Tuple
+from typing import Tuple, Optional
 
 import arff
 import pandas as pd
@@ -19,6 +19,9 @@ def arff_to_pandas(file_path: str) -> pd.DataFrame:
         A dataframe of the data in the ARFF file,
         with categorical columns having category dtype.
     """
+    if not isinstance(file_path, str):
+        raise TypeError(f"`file_path` must be of type `str` but is of type {type(file_path)}")
+
     with open(file_path, 'r') as arff_file:
         arff_dict = arff.load(arff_file)
 
@@ -31,17 +34,17 @@ def arff_to_pandas(file_path: str) -> pd.DataFrame:
     return data
 
 
-def X_y_from_arff(file_path: str, split_column: str = 'last') -> Tuple[pd.DataFrame, pd.Series]:
+def X_y_from_arff(file_path: str, split_column: Optional[str] = None) -> Tuple[pd.DataFrame, pd.Series]:
     """ Load data from the ARFF file into pandas DataFrame and specified column to pd.Series. "
 
     Parameters
     ----------
     file_path: str
         path to the ARFF file.
-    split_column: str (default='last')
+    split_column: str, optional (default=None)
         Column to split and return separately (e.g. target column).
-        Value should either match a column name or 'last'.
-        If 'last' is specified, the last column is returned separately.
+        Value should either match a column name or None.
+        If None is specified, the last column is returned separately.
 
     Returns
     -------
@@ -50,7 +53,7 @@ def X_y_from_arff(file_path: str, split_column: str = 'last') -> Tuple[pd.DataFr
     """
     data = arff_to_pandas(file_path)
 
-    if split_column == 'last':
+    if split_column is None:
         return data.iloc[:, :-1], data.iloc[:, -1]
     elif split_column in data.columns:
         return data.loc[:, data.columns != split_column], data.loc[:, split_column]

gama/gama.py

@@ -21,6 +21,7 @@
 from gama.utilities.metrics import scoring_to_metric
 from .utilities.observer import Observer
 
+from gama.__version__ import __version__
 from gama.data import X_y_from_arff
 from gama.search_methods.async_ea import AsyncEA
 from gama.utilities.generic.timekeeper import TimeKeeper
@@ -42,7 +43,6 @@
 STR_NO_OPTIMAL_PIPELINE = """Gama did not yet establish an optimal pipeline.
                           This can be because `fit` was not yet called, or
                           did not terminate successfully."""
-__version__ = '19.01.0'
 
 for module_to_ignore in ["sklearn", "numpy"]:
     warnings.filterwarnings("ignore", module=module_to_ignore)
@@ -202,23 +202,24 @@ def predict(self, x: Union[pd.DataFrame, np.ndarray]):
                 x[col] = x[col].astype(self._X[col].dtype)
         return self._predict(x)
 
-    def predict_arff(self, arff_file_path: str):
+    def predict_arff(self, arff_file_path: str, target_column: Optional[str] = None) -> np.ndarray:
         """ Predict the target for input found in the ARFF file.
 
         Parameters
         ----------
         arff_file_path: str
             An ARFF file with the same columns as the one that used in fit.
-            The target column is ignored (but must be present).
+            Target column must be present in file, but its values are ignored (can be '?').
+        target_column: str, optional (default=None)
+            Specifies which column the model should predict.
+            If left None, the last column is taken to be the target.
 
         Returns
         -------
         numpy.ndarray
             array with predictions for each row in the ARFF file.
         """
-        if not isinstance(arff_file_path, str):
-            raise TypeError(f"`arff_file_path` must be of type `str` but is of type {type(arff_file_path)}")
-        X, _ = X_y_from_arff(arff_file_path)
+        X, _ = X_y_from_arff(arff_file_path, split_column=target_column)
         return self._predict(X)
 
     def score(self, x: Union[pd.DataFrame, np.ndarray], y: Union[pd.Series, np.ndarray]) -> float:
@@ -239,32 +240,38 @@ def score(self, x: Union[pd.DataFrame, np.ndarray], y: Union[pd.Series, np.ndarr
         predictions = self.predict_proba(x) if self._metrics[0].requires_probabilities else self.predict(x)
         return self._metrics[0].score(y, predictions)
 
-    def score_arff(self, arff_file_path: str) -> float:
+    def score_arff(self, arff_file_path: str, target_column: Optional[str] = None) -> float:
         """ Calculate the score of the model according to the `scoring` metric and input in the ARFF file.
 
         Parameters
         ----------
-        arff_file_path: string
+        arff_file_path: str
             An ARFF file with which to calculate the score.
+        target_column: str, optional (default=None)
+            Specifies which column the model should predict.
+            If left None, the last column is taken to be the target.
 
         Returns
         -------
         float
             The score obtained on the given test data according to the `scoring` metric.
         """
-        X, y = X_y_from_arff(arff_file_path)
+        X, y = X_y_from_arff(arff_file_path, split_column=target_column)
         return self.score(X, y)
 
-    def fit_arff(self, arff_file_path: str, *args, **kwargs):
+    def fit_arff(self, arff_file_path: str, target_column: Optional[str] = None, *args, **kwargs):
         """ Find and fit a model to predict the target column (last) from other columns.
 
         Parameters
         ----------
-        arff_file_path: string
+        arff_file_path: str
             Path to an ARFF file containing the training data.
-            The last column is always taken to be the target.
+        target_column: str, optional (default=None)
+            Specifies which column the model should predict.
+            If left None, the last column is taken to be the target.
+
         """
-        X, y = X_y_from_arff(arff_file_path)
+        X, y = X_y_from_arff(arff_file_path, split_column=target_column)
         self.fit(X, y, *args, **kwargs)
 
     def fit(self,

gama/utilities/preprocessing.py

@@ -86,8 +86,12 @@ def format_x_y(x: Union[pd.DataFrame, np.ndarray], y: Union[pd.DataFrame, pd.Ser
 
     if isinstance(x, np.ndarray):
         x = heuristic_numpy_to_dataframe(x)
-    if isinstance(y, np.ndarray) and y.ndim == 2 and y.shape[1] > 1:
-        y = np.argmax(y, axis=1)
+    if isinstance(y, np.ndarray) and y.ndim == 2:
+        # Either indicator matrix or should be a vector.
+        if y.shape[1] > 1:
+            y = np.argmax(y, axis=1)
+        else:
+            y = y.squeeze()
 
     if y_type == pd.Series:
         if isinstance(y, pd.DataFrame):

setup.py

@@ -3,6 +3,9 @@
 
 from setuptools import setup, find_packages
 
+with open("gama/__version__.py", 'r') as fh:
+    version = fh.readlines()[-1].split()[-1].strip("\"'")
+
 requirements = [
     'numpy>=1.14.0',
     'scipy>=1.0.0',
@@ -37,7 +40,7 @@
 
 setup(
     name='gama',
-    version='19.08.0',
+    version=version,
     description='A package for automated machine learning based on scikit-learn.',
     long_description=README,
     long_description_content_type='text/markdown',

tests/unit/test_preprocessing.py

@@ -14,7 +14,8 @@ def well_formatted_x_y(x, y, y_type):
     X_np, y_np = load_digits(return_X_y=True)
     X_df, y_df = pd.DataFrame(X_np), pd.DataFrame(y_np)
     y_series = pd.Series(y_np)
+    y_2d = y_np.reshape(-1, 1)
 
-    for X, y in itertools.product([X_np, X_df], [y_np, y_series, y_df]):
+    for X, y in itertools.product([X_np, X_df], [y_np, y_series, y_df, y_2d]):
         well_formatted_x_y(*format_x_y(X, y), y_type=pd.Series)
         well_formatted_x_y(*format_x_y(X, y, y_type=pd.DataFrame), y_type=pd.DataFrame)