DemoRegression.ipynb reveals some bug

[gciatto]

The following cell in demos/DemoRegression.ipynb:

crash = CRASH(predictor, train, max_depth=3, patience=1, readability_tradeoff=.75, algorithm=CRASH.Algorithm.ORCHiD)
crash.search()
(_, _, depth, threshold) = crash.get_best()[0]

orchid = Extractor.orchid(predictor, depth=depth, error_threshold=threshold, output=Target.REGRESSION)
theory_from_orchid = orchid.extract(train)
evaluate('ORCHiD', orchid, true, predicted)
print('ORCHiD extracted rules:\n\n' + pretty_theory(theory_from_orchid))

produces the following output and error:

Algorithm.ORCHiD. Depth: 1. Threshold = 0.00. MAE = 10.29, 2 rules
Algorithm.ORCHiD. Depth: 1. Threshold = 0.00. MAE = 8.47, 2 rules
Algorithm.ORCHiD. Depth: 1. Threshold = 0.00. MAE = 10.39, 2 rules

Algorithm.ORCHiD. Depth: 2. Threshold = 0.00. MAE = 6.76, 4 rules
Algorithm.ORCHiD. Depth: 2. Threshold = 0.00. MAE = 6.06, 4 rules
Algorithm.ORCHiD. Depth: 2. Threshold = 0.00. MAE = 6.71, 4 rules

Algorithm.ORCHiD. Depth: 3. Threshold = 0.00. MAE = 5.66, 8 rules
Algorithm.ORCHiD. Depth: 3. Threshold = 0.00. 

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In [10], line 2
      1 crash = CRASH(predictor, train, max_depth=3, patience=1, readability_tradeoff=.75, algorithm=CRASH.Algorithm.ORCHiD)
----> 2 crash.search()
      3 (_, _, depth, threshold) = crash.get_best()[0]
      5 orchid = Extractor.orchid(predictor, depth=depth, error_threshold=threshold, output=Target.REGRESSION)

File /usr/local/lib/python3.9/site-packages/psyke/tuning/crash/__init__.py:25, in CRASH.search(self)
     24 def search(self):
---> 25     self.params = self.__search_depth()

File /usr/local/lib/python3.9/site-packages/psyke/tuning/crash/__init__.py:32, in CRASH.__search_depth(self)
     29 best = None
     31 for depth in range(1, self.max_depth + 1):
---> 32     p = self.__search_threshold(depth)
     33     b = Optimizer._best(p)[1]
     34     print()

File /usr/local/lib/python3.9/site-packages/psyke/tuning/crash/__init__.py:57, in CRASH.__search_threshold(self, depth)
     52 print(f"{self.algorithm}. Depth: {depth}. Threshold = {threshold:.2f}. ", end="")
     53 extractor = Extractor.creepy(self.predictor, depth, threshold, self.output, 10,
     54                              normalization=self.normalization) if self.algorithm == CRASH.Algorithm.CReEPy \
     55     else Extractor.orchid(self.predictor, depth, threshold, self.output, 10,
     56                           normalization=self.normalization)
---> 57 _ = extractor.extract(self.dataframe)
     58 mae, n = (extractor.mae(self.dataframe, self.predictor) if self.objective == Objective.MODEL else
     59           extractor.mae(self.dataframe)), extractor.n_rules
     60 print(f"MAE = {mae:.2f}, {n} rules")

File /usr/local/lib/python3.9/site-packages/psyke/__init__.py:251, in PedagogicalExtractor.extract(self, dataframe, mapping)
    249 data = dataframe.iloc[:, :-1].copy().join(new_y)
    250 data.columns = dataframe.columns
--> 251 return self._extract(data, mapping)

File /usr/local/lib/python3.9/site-packages/psyke/extraction/hypercubic/creepy/__init__.py:28, in CReEPy._extract(self, dataframe, mapping)
     27 def _extract(self, dataframe: pd.DataFrame, mapping: dict[str: int] = None) -> Theory:
---> 28     self._hypercubes = self.clustering.extract(dataframe)
     29     for cube in self._hypercubes:
     30         for dimension in self._ignore_dimensions():

File /usr/local/lib/python3.9/site-packages/psyke/clustering/exact/__init__.py:55, in ExACT.extract(self, dataframe)
     52 def extract(self, dataframe: pd.DataFrame) -> Iterable[HyperCube]:
     53     self._predictor.fit(dataframe.iloc[:, :-1], dataframe.iloc[:, -1])
     54     self._hypercubes = \
---> 55         self._iterate(Node(dataframe, HyperCube.create_surrounding_cube(dataframe, True, self._output)))
     56     return list(self._hypercubes)

File /usr/local/lib/python3.9/site-packages/psyke/clustering/cream/__init__.py:54, in CREAM._iterate(self, surrounding)
     52 gauss_params = select_gaussian_mixture(data, self.gauss_components)
     53 gauss_pred = gauss_params[2].predict(data)
---> 54 cubes = self.__eligible_cubes(gauss_pred, node, gauss_params[1])
     55 if len(cubes) < 1:
     56     continue

File /usr/local/lib/python3.9/site-packages/psyke/clustering/cream/__init__.py:29, in CREAM.__eligible_cubes(self, gauss_pred, node, clusters)
     27 if len(df) == 0:
     28     continue
---> 29 inner_cube = self._create_cube(df, clusters)
     30 indices = self._indices(inner_cube, node.dataframe)
     31 if indices is None:

File /usr/local/lib/python3.9/site-packages/psyke/clustering/exact/__init__.py:46, in ExACT._create_cube(self, dataframe, clusters)
     44 def _create_cube(self, dataframe: pd.DataFrame, clusters: int) -> ClosedCube:
     45     data = ExACT._remove_string_label(dataframe)
---> 46     dbscan_pred = DBSCAN(eps=select_dbscan_epsilon(data, clusters)).fit_predict(data.iloc[:, :-1])
     47     return HyperCube.create_surrounding_cube(
     48         dataframe.iloc[np.where(dbscan_pred == Counter(dbscan_pred).most_common(1)[0][0])],
     49         True, self._output
     50     )

File /usr/local/lib/python3.9/site-packages/psyke/clustering/utils.py:32, in select_dbscan_epsilon(data, clusters)
     30     epsilon = max(distances[-1], 1e-3)
     31 k = 1.
---> 32 dbscan_pred = DBSCAN(eps=epsilon * k).fit_predict(data.iloc[:, :-1])
     33 # while Counter(dbscan_pred).most_common(1)[0][0] == -1:
     34 for i in range(1000):

TypeError: unsupported operand type(s) for *: 'NoneType' and 'float'