Merge pull request #64 from jni/fix-learning-mode

Fix #63

.travis.yml

@@ -14,7 +14,7 @@ install:
   - conda update -q conda
   - conda info -a
 
-  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION numpy scipy matplotlib networkx cython h5py pillow scikit-image scikit-learn setuptools pip
+  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION numpy scipy matplotlib networkx cython h5py pillow scikit-image scikit-learn=0.16 setuptools pip
   - source activate test-environment
 
   # custom package not available from conda

gala/agglo.py

@@ -39,7 +39,7 @@
 def contingency_table(a, b):
     ct = ev_contingency_table(a, b)
     nx, ny = ct.shape
-    ctout = np.zeros((2 * nx, ny), ct.dtype)
+    ctout = np.zeros((2*nx + 1, ny), ct.dtype)
     ct.todense(out=ctout[:nx, :])
     return ctout
 
@@ -202,7 +202,7 @@ def classifier_probability(feature_extractor, classifier):
     def predict(g, n1, n2):
         if n1 == g.boundary_body or n2 == g.boundary_body:
             return inf
-        features = feature_extractor(g, n1, n2)
+        features = np.atleast_2d(feature_extractor(g, n1, n2))
         try:
             prediction = classifier.predict_proba(features)
             prediction_arr = np.array(prediction, copy=False)
@@ -616,9 +616,6 @@ def set_probabilities(self, probs=array([]), normalize=False):
             self.probabilities = morpho.pad(probs, padding)
             self.probabilities_r = self.probabilities.ravel()[:,newaxis]
         elif p_ndim == w_ndim+1:
-            if sp[1:] == sw:
-                sp = sp[1:]+[sp[0]]
-                probs = probs.transpose(sp)
             axes = list(range(p_ndim-1))
             self.probabilities = morpho.pad(probs, padding, axes)
             self.probabilities_r = self.probabilities.reshape(
@@ -1087,10 +1084,7 @@ def learn_agglomerate(self, gts, feature_map,
                 g.merge_priority_function = boundary_mean
             elif num_epochs > 0 and priority_mode == 'active' or \
                 num_epochs % 2 == 1 and priority_mode == 'mixed':
-                if random_state == None:
-                    cl = get_classifier(classifier)
-                else:
-                    cl = get_classifier(classifier, random_state=random_state)
+                cl = get_classifier(classifier, random_state=random_state)
                 feat, lab = classify.sample_training_data(
                     data[0], data[1][:, label_type_keys[labeling_mode]],
                     max_num_samples)
@@ -1104,8 +1098,9 @@ def learn_agglomerate(self, gts, feature_map,
             g.show_progress = False # bug in MergeQueue usage causes
                                     # progressbar crash.
             g.rebuild_merge_queue()
-            alldata.append(g._learn_agglomerate(ctables, feature_map,
-                                                learning_mode, labeling_mode))
+            alldata.append(g.learn_epoch(ctables, feature_map,
+                                         learning_mode=learning_mode,
+                                         labeling_mode=labeling_mode))
             if memory:
                 if unique:
                     data = unique_learning_data_elements(alldata)
@@ -1210,8 +1205,8 @@ def learn_edge(self, edge, ctables, assignments, feature_map):
         return features, labels, weights, (n1, n2)
 
 
-    def _learn_agglomerate(self, ctables, feature_map, gt_dts,
-                        learning_mode='strict', labeling_mode='assignment'):
+    def learn_epoch(self, ctables, feature_map,
+                    learning_mode='permissive', labeling_mode='assignment'):
         """Learn the agglomeration process using various strategies.
 
         Parameters
@@ -1223,10 +1218,11 @@ def _learn_agglomerate(self, ctables, feature_map, gt_dts,
             The map from node pairs to a feature vector. This must
             consist either of uncached features or of the cache used
             when building the graph.
-        learning_mode : {'strict', 'loose'}
+        learning_mode : {'strict', 'permissive'}, optional
             If ``'strict'``, don't proceed with a merge when it goes against
-            the ground truth.
-        labeling_mode : {'assignment', 'vi-sign', 'rand-sign'}
+            the ground truth. For historical reasons, 'loose' is allowed as
+            a synonym for 'strict'.
+        labeling_mode : {'assignment', 'vi-sign', 'rand-sign'}, optional
             Which label to use for `learning_mode`. Note that all labels
             are saved in the end.
 
@@ -1682,12 +1678,7 @@ def traversing_bodies(self):
     def non_traversing_bodies(self):
         """List bodies that are not orphans and do not traverse the volume."""
         return [n for n in self.nodes() if self.at_volume_boundary(n) and
-            not self.is_traversed_by_node(n)]
-
-
-    def compute_non_traversing_bodies(self):
-        """Same as agglo.Rag.non_traversing_bodies, but doesn't use graph."""
-        return morpho.non_traversing_bodies(self.get_segmentation())
+                not self.is_traversed_by_node(n) and n != self.boundary_body]
 
 
     def raveler_body_annotations(self, traverse=False):

gala/classify.py

@@ -9,11 +9,8 @@
 # libraries
 import h5py
 import numpy as np
-from numpy.testing import assert_raises
 np.seterr(divide='ignore')
 
-from sklearn.svm import SVC
-from sklearn.linear_model import LogisticRegression, LinearRegression
 from sklearn.ensemble import RandomForestClassifier
 from sklearn.externals import joblib
 
@@ -26,9 +23,6 @@
 else:
     vigra_available = True
 
-# local imports
-from . import iterprogress as ip
-
 
 def default_classifier_extension(cl, use_joblib=True):
     """
@@ -166,6 +160,7 @@ def get_classifier(name='random forest', *args, **kwargs):
     True
     >>> cl.n_estimators
     47
+    >>> from numpy.testing import assert_raises
     >>> assert_raises(NotImplementedError, get_classifier, 'perfect class')
     """
     name = name.lower()
@@ -177,6 +172,8 @@ def get_classifier(name='random forest', *args, **kwargs):
         return DefaultRandomForest(*args, **kwargs)
     elif is_naive_bayes:
         from sklearn.naive_bayes import GaussianNB
+        if 'random_state' in kwargs:
+            del kwargs['random_state']
         return GaussianNB(*args, **kwargs)
     else:
         raise NotImplementedError('Classifier "%s" is either not installed '

gala/imio.py

@@ -280,7 +280,7 @@ def write_png_image_stack(npy_vol, fn, axis=-1, bitdepth=None):
         npy_vol = uint32(npy_vol)
     for z, pl in enumerate(npy_vol):
         im = Image.new(mode_base, pl.T.shape)
-        im.fromstring(pl.tostring(), 'raw', mode)
+        im.frombytes(pl.tostring(), 'raw', mode)
         im.save(fn % z)
 
 ### VTK structured points array format

tests/test_agglo.py

@@ -98,6 +98,15 @@ def test_mask():
     assert (2, 4) in g.edges()
 
 
+def test_traverse():
+    labels = [[0, 1, 2],
+              [0, 1, 2],
+              [0, 1, 2]]
+    g = agglo.Rag(np.array(labels))
+    assert g.traversing_bodies() == [1]
+    assert g.non_traversing_bodies() == [0, 2]
+
+
 if __name__ == '__main__':
     from numpy import testing
     testing.run_module_suite()

tests/test_gala.py

@@ -1,5 +1,4 @@
 import os
-import sys
 import glob
 from contextlib import contextmanager
 
@@ -67,7 +66,7 @@ def train_and_save_classifier(training_data_file, filename,
 
 ### tests
 
-def test_generate_examples_1_channel():
+def test_generate_lash_examples_1_channel():
     """Run a flat epoch and an active epoch of learning, compare learned sets.
 
     The *order* of the edges learned by learn_flat is not guaranteed, so we
@@ -78,6 +77,7 @@ def test_generate_examples_1_channel():
     """
     g_train = agglo.Rag(ws_train, pr_train, feature_manager=fc)
     _, alldata = g_train.learn_agglomerate(gt_train, fc,
+                                           learning_mode='permissive',
                                            classifier='naive bayes')
     testfn = 'example-data/train-naive-bayes-merges1-py3.pck'
     exp0, exp1 = load_pickle(os.path.join(rundir, testfn))
@@ -94,6 +94,24 @@ def test_generate_examples_1_channel():
     assert_allclose(nb.class_prior_, nbexp.class_prior_, atol=1e-7)
 
 
+def test_generate_gala_examples_1_channel():
+    """As `test_generate_lash_examples_1_channel`, but using strict learning.
+    """
+    g_train = agglo.Rag(ws_train, pr_train, feature_manager=fc)
+    _, alldata = g_train.learn_agglomerate(gt_train, fc,
+                                           learning_mode='strict',
+                                           classifier='naive bayes')
+    testfn = 'example-data/train-naive-bayes-merges1-py3.pck'
+    exp0, exp1 = load_pickle(os.path.join(rundir, testfn))
+    expected_edges = set(map(tuple, exp0))
+    edges = set(map(tuple, alldata[0][3]))
+    merges = alldata[1][3]
+    # expect same edges in flat learning
+    assert edges == expected_edges  # flat learning epoch
+    # expect more edges in strict training than permissive
+    assert np.shape(merges)[0] > np.shape(exp1)[0]
+
+
 def test_segment_with_classifer_1_channel():
     fn = os.path.join(rundir, 'example-data/rf1-py3.joblib')
     with tar_extract(fn) as fn:
@@ -119,6 +137,7 @@ def test_generate_examples_4_channel():
     """
     g_train = agglo.Rag(ws_train, p4_train, feature_manager=fc)
     _, alldata = g_train.learn_agglomerate(gt_train, fc,
+                                           learning_mode='permissive',
                                            classifier='naive bayes')
     testfn = 'example-data/train-naive-bayes-merges4-py3.pck'
     exp0, exp1 = load_pickle(os.path.join(rundir, testfn))