BUG: fix GaussianHMM.fit to allow input sequences of different lengths

ronw · Fabian Pedregosa · commit d1c6e5a65745 · 2011-02-23T18:01:21.000+01:00
diff --git a/scikits/learn/hmm.py b/scikits/learn/hmm.py
@@ -327,8 +327,6 @@ def fit(self, obs, n_iter=10, thresh=1e-2, params=string.letters,
         small).  You can fix this by getting more training data, or
         decreasing `covars_prior`.
         """
-        obs = np.asanyarray(obs)
-
         self._init(obs, init_params)
 
         logprob = []
@@ -679,11 +677,13 @@ def _generate_sample_from_state(self, state):
     def _init(self, obs, params='stmc'):
         super(GaussianHMM, self)._init(obs, params=params)
 
-        if hasattr(self, 'n_features') and self.n_features != obs.shape[2]:
+        if (hasattr(self, 'n_features')
+            and self.n_features != obs[0].shape[1]):
             raise ValueError('Unexpected number of dimensions, got %s but '
-                             'expected %s' % (obs.shape[2], self.n_features))
+                             'expected %s' % (obs[0].shape[1],
+                                              self.n_features))
 
-        self.n_features = obs.shape[2]
+        self.n_features = obs[0].shape[1]
 
         if 'm' in params:
             self._means = cluster.KMeans(
diff --git a/scikits/learn/tests/test_hmm.py b/scikits/learn/tests/test_hmm.py
@@ -330,6 +330,16 @@ def test_fit(self, params='stmc', n_iter=15, verbose=False, **kwargs):
                    % (self.cvtype, params, trainll, np.diff(trainll)))
         self.assertTrue(np.all(np.diff(trainll) > -0.5))
 
+    def test_fit_works_on_sequences_of_different_length(self):
+        obs = [np.random.rand(3, self.n_features),
+               np.random.rand(4, self.n_features),
+               np.random.rand(5, self.n_features)]
+
+        h = hmm.GaussianHMM(self.n_states, self.cvtype)
+        # This shouldn't raise
+        # ValueError: setting an array element with a sequence.
+        h.fit(obs)
+
     def test_fit_with_priors(self, params='stmc', n_iter=10,
                              verbose=False):
         startprob_prior = 10 * self.startprob + 2.0
@@ -612,6 +622,16 @@ def test_fit(self, params='stmwc', n_iter=5, verbose=True, **kwargs):
                                                     np.diff(trainll))
         self.assertTrue(np.all(np.diff(trainll) > -0.5))
 
+    def test_fit_works_on_sequences_of_different_length(self):
+        obs = [np.random.rand(3, self.n_features),
+               np.random.rand(4, self.n_features),
+               np.random.rand(5, self.n_features)]
+
+        h = hmm.GMMHMM(self.n_states, cvtype=self.cvtype)
+        # This shouldn't raise
+        # ValueError: setting an array element with a sequence.
+        h.fit(obs)
+
 
 class TestGMMHMMWithSphericalCovars(TestGMMHMM):
     cvtype = 'spherical'
diff --git a/scikits/learn/tests/test_mixture.py b/scikits/learn/tests/test_mixture.py
@@ -169,6 +169,18 @@ def test_GMM_attributes():
     assert_raises(ValueError, mixture.GMM, n_states=20, cvtype='badcvtype')
 
 
+def test_GMM_fit_works_on_sequences_of_different_length():
+    ndim = 3
+    obs = [np.random.rand(3, ndim),
+           np.random.rand(4, ndim),
+           np.random.rand(5, ndim)]
+
+    gmm = mixture.GMM(n_states=1)
+    # This shouldn't raise
+    # ValueError: setting an array element with a sequence.
+    gmm.fit(obs)
+
+
 class GMMTester():
     n_states = 10
     n_features = 4
