Merge pull request #531 from markovmodel/rename_transformer_functions

Rename transformer functions

pyemma/_base/estimator.py

@@ -143,8 +143,6 @@ def _estimate_param_scan_worker(estimator, params, X, evaluate, evaluate_args,
 
 def estimate_param_scan(estimator, X, param_sets, evaluate=None, evaluate_args=None, failfast=True,
                         return_estimators=False, n_jobs=1, progress_reporter=None):
-    # TODO: parallelize. For options see http://scikit-learn.org/stable/modules/grid_search.html
-    # TODO: allow to specify method parameters in evaluate
     """ Runs multiple estimations using a list of parameter settings
 
     Parameters

pyemma/coordinates/api.py

@@ -935,7 +935,7 @@ def pca(data=None, dim=2, var_cutoff=0.95, stride=1, mean=None):
         indim = data.dimension()
         mean = _types.ensure_ndarray(mean, shape=(indim,), dtype=_np.float)
 
-    res = _PCA(dim=dim, var_cutoff=var_cutoff)
+    res = _PCA(dim=dim, var_cutoff=var_cutoff, mean=mean)
     return _param_stage(data, res, stride=stride)
 
 

pyemma/coordinates/clustering/assign.py

@@ -99,7 +99,7 @@ def _param_add_data(self, X, itraj, t, first_chunk, last_chunk_in_traj,
             self._dtrajs.append(np.empty(n, dtype=self.output_type()))
 
         L = np.shape(X)[0]
-        self._dtrajs[itraj][t:t+L] = self._map_array(X).squeeze()
+        self._dtrajs[itraj][t:t+L] = self._transform_array(X).squeeze()
 
         if last_chunk:
             return True

pyemma/coordinates/clustering/interface.py

@@ -128,7 +128,7 @@ def sample_indexes_by_cluster(self, clusters, nsample, replace=True):
 
         return sample_indexes_by_state(self._index_states[clusters], nsample, replace=replace)
 
-    def _map_array(self, X):
+    def _transform_array(self, X):
         """get closest index of point in :attr:`clustercenters` to x."""
         dtraj = np.empty(X.shape[0], dtype=self.output_type())
         regspatial.assign(X.astype(np.float32, order='C', copy=False),
@@ -192,7 +192,7 @@ def assign(self, X=None, stride=1):
                 raise ValueError('assign accepts either X or stride parameters, but not both. If you want to map '+
                                  'only a subset of your data, extract the subset yourself and pass it as X.')
             # map to column vector(s)
-            mapped = self.map(X)
+            mapped = self.transform(X)
             # flatten
             if isinstance(mapped, np.ndarray):
                 mapped = np.transpose(mapped)[0]

pyemma/coordinates/data/feature_reader.py

@@ -251,16 +251,16 @@ def _next_chunk(self, context=None):
                 shape_2d = (shape[0], shape[1] * shape[2])
                 return chunk.xyz.reshape(shape_2d)
             else:
-                return self.featurizer.map(chunk)
+                return self.featurizer.transform(chunk)
         else:
             if len(self.featurizer.active_features) == 0:
                 shape_Y = adv_chunk.xyz.shape
 
                 X = chunk.xyz.reshape((shape[0], shape[1] * shape[2]))
                 Y = adv_chunk.xyz.reshape((shape_Y[0], shape_Y[1] * shape_Y[2]))
             else:
-                X = self.featurizer.map(chunk)
-                Y = self.featurizer.map(adv_chunk)
+                X = self.featurizer.transform(chunk)
+                Y = self.featurizer.transform(adv_chunk)
             return X, Y
 
     def parametrize(self, stride=1):

pyemma/coordinates/data/featurizer.py

@@ -32,14 +32,13 @@
 import warnings
 from itertools import combinations as _combinations, count
 from itertools import product as _product
-from pyemma.util.types import is_iterable_of_int as _is_iterable_of_int
 import functools
 
 
 from six import PY3
 from pyemma.util.types import is_iterable_of_int as _is_iterable_of_int
 from pyemma.util.log import getLogger
-#from pyemma.util.annotators import deprecated
+from pyemma.util.annotators import deprecated
 from six.moves import map
 from six.moves import range
 from six.moves import zip
@@ -285,7 +284,14 @@ def describe(self):
                                                            str(self._args) +
                                                            str(self._kwargs))]
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         feature = self._func(traj, *self._args, **self._kwargs)
         if not isinstance(feature, np.ndarray):
             raise ValueError("your function should return a NumPy array!")
@@ -332,7 +338,14 @@ def describe(self):
     def dimension(self):
         return 3 * self.indexes.shape[0]
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         newshape = (traj.xyz.shape[0], 3 * self.indexes.shape[0])
         return np.reshape(traj.xyz[:, self.indexes, :], newshape)
 
@@ -366,7 +379,14 @@ def describe(self):
     def dimension(self):
         return self.distance_indexes.shape[0]
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         return mdtraj.compute_distances(traj, self.distance_indexes, periodic=self.periodic)
 
     def __hash__(self):
@@ -386,7 +406,14 @@ def __init__(self, top, distance_indexes, periodic=True):
             self, top, distance_indexes, periodic=periodic)
         self.prefix_label = "INVDIST:"
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         return 1.0 / mdtraj.compute_distances(traj, self.distance_indexes, periodic=self.periodic)
 
     # does not need own hash impl, since we take prefix label into account
@@ -417,7 +444,14 @@ def describe(self):
                   for pair in self.distance_indexes]
         return labels
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         # We let mdtraj compute the contacts with the input scheme
         D = mdtraj.compute_contacts(traj, contacts=self.contacts, scheme=self.scheme)[0]
         res = np.zeros_like(D)
@@ -446,7 +480,14 @@ def describe(self):
                   for pair in self.distance_indexes]
         return labels
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         # All needed distances
         Dall = mdtraj.compute_distances(traj, self.distance_list)
         # Just the minimas
@@ -472,7 +513,14 @@ def __init__(self, top, distance_indexes, threshold=5.0, periodic=True):
         self.threshold = threshold
         self.periodic = periodic
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         dists = mdtraj.compute_distances(
             traj, self.distance_indexes, periodic=self.periodic)
         res = np.zeros(
@@ -496,7 +544,6 @@ def __init__(self, top, angle_indexes, deg=False, cossin=False):
         self.cossin = cossin
 
     def describe(self):
-
         if self.cossin:
             sin_cos = ("ANGLE: COS(%s - %s - %s)",
                        "ANGLE: SIN(%s - %s - %s)")
@@ -520,7 +567,14 @@ def dimension(self):
             dim *= 2
         return dim
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         rad = mdtraj.compute_angles(traj, self.angle_indexes)
         if self.cossin:
             rad = np.dstack((np.cos(rad), np.sin(rad)))
@@ -576,7 +630,14 @@ def describe(self):
     def dimension(self):
         return self._dim
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         rad = mdtraj.compute_dihedrals(traj, self.dih_indexes)
         if self.cossin:
             rad = np.dstack((np.cos(rad), np.sin(rad)))
@@ -719,7 +780,14 @@ def describe(self):
     def dimension(self):
         return 1
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         return np.array(mdtraj.rmsd(traj, self.ref, atom_indices=self.atom_indices), ndmin=2).T
 
     def __hash__(self):
@@ -1298,7 +1366,14 @@ def dimension(self):
         dim = sum(f.dimension for f in self.active_features)
         return dim
 
+    @deprecated
     def map(self, traj):
+        r"""Deprecated: use transform(traj)
+
+        """
+        return self.transform(traj)
+
+    def transform(self, traj):
         """
         Maps an mdtraj Trajectory object to the selected output features
 
@@ -1337,7 +1412,7 @@ def map(self, traj):
             # perform sanity checks for custom feature input
             if isinstance(f, CustomFeature):
                 # NOTE: casting=safe raises in numpy>=1.9
-                vec = f.map(traj).astype(np.float32, casting='safe')
+                vec = f.transform(traj).astype(np.float32, casting='safe')
                 if vec.shape[0] == 0:
                     vec = np.empty((0, f.dimension))
 
@@ -1357,7 +1432,7 @@ def map(self, traj):
                                         traj.xyz.shape[0],
                                         vec.shape[0]))
             else:
-                vec = f.map(traj).astype(np.float32)
+                vec = f.transform(traj).astype(np.float32)
             feature_vec.append(vec)
 
         if len(feature_vec) > 1:

pyemma/coordinates/data/interface.py

@@ -156,10 +156,10 @@ def _add_array_to_storage(self, array):
         self._data.append(array)
 
     # handle abstract methods and special cases
-    def map(self, X):
+    def transform(self, X):
         raise NotImplementedError("a reader can not map data, it is a data source")
 
-    def _map_array(self, X):
+    def _transform_array(self, X):
         raise NotImplementedError("a reader can not map data, it is a data source")
 
     def _param_add_data(self, *args, **kwargs):

pyemma/coordinates/data/writer.py

@@ -59,7 +59,7 @@ def describe(self):
     def dimension(self):
         return self.data_producer.dimension()
 
-    def _map_array(self, X):
+    def _transform_array(self, X):
         pass
 
     def _reset(self, stride=1):

pyemma/coordinates/pipelines.py

@@ -140,19 +140,7 @@ def set_element(self, index, e):
 
         return replaced
 
-    def run(self):
-        """deprecated. Identical to parametrize()
-        """
-        import warnings
-        warnings.warn(
-            "run() is deprecated and will be disabled in the future. Use parametrize().", DeprecationWarning)
-        self.parametrize()
-
-    # TODO: DISCUSS - renamed run() to parametrize (because run is a bit ambiguous).
-    # TODO: We could also call it fit() (here and in the transformers).
-    # TODO: This might be nicer because it's shorter and the spelling is unambiguous
-    # TODO: (in contrast to parametrize and parameterize and parameterise that
-    # are all correct in english.
+    # TODO: to be replaced by fit/estimate
     def parametrize(self):
         r"""
         Reads all data and discretizes it into discrete trajectories.

pyemma/coordinates/tests/test_assign.py

@@ -132,7 +132,7 @@ def test_iterator(self):
 
     def test_map(self):
         c = self.ass
-        Y = c.map(self.X)
+        Y = c.transform(self.X)
         assert Y.shape[0] == self.T
         assert Y.shape[1] == 1
         # test if consistent with get_output

pyemma/coordinates/tests/test_cluster.py

@@ -126,7 +126,7 @@ def test_iterator(self):
 
     def test_map(self):
         for c in self.cl:
-            Y = c.map(self.X)
+            Y = c.transform(self.X)
             assert Y.shape[0] == self.T
             assert Y.shape[1] == 1
             # test if consistent with get_output