modAL.models.ActiveLearner teach method extended, now it is possible to train only on new data

Author: cosmic-cortex

modAL/cluster.py

@@ -9,9 +9,9 @@
 
 
 class HierarchicalClustering:
-    def __init__(self, X, labels):
-        self.labels = labels
-        self.cluster = AgglomerativeClustering(n_clusters=2, compute_full_tree=True)
+    def __init__(self, X, classes, n_batch=1):
+        self.classes = classes
+        self.cluster = AgglomerativeClustering(compute_full_tree=True)
         self.cluster.fit(X)
 
     def __call__(self, *args, **kwargs):

modAL/models.py

@@ -320,7 +320,7 @@ def score(self, X, y, **score_kwargs):
         """
         return self.estimator.score(X, y, **score_kwargs)
 
-    def teach(self, X, y, bootstrap=False, **fit_kwargs):
+    def teach(self, X, y, bootstrap=False, only_new=False, **fit_kwargs):
         """
         Adds X and y to the known training data and retrains the predictor
         with the augmented dataset.
@@ -338,12 +338,20 @@ def teach(self, X, y, bootstrap=False, **fit_kwargs):
             If True, training is done on a bootstrapped dataset. Useful for building
             Committee models with bagging.
 
+        only_new: boolean
+            If True, the model is retrained using only X and y, ignoring the previously
+            provided examples. Useful when working with models where the .fit() method
+            doesn't retrain the model from scratch. (For example, in tensorflow or keras.)
+
         fit_kwargs: keyword arguments
             Keyword arguments to be passed to the fit method
             of the predictor.
         """
         self._add_training_data(X, y)
-        self._fit_to_known(bootstrap=bootstrap, **fit_kwargs)
+        if not only_new:
+            self._fit_to_known(bootstrap=bootstrap, **fit_kwargs)
+        else:
+            self._fit_on_new(X, y, bootstrap=bootstrap, **fit_kwargs)
 
 
 class BaseCommittee(ABC, BaseEstimator):

tests/core_tests.py

@@ -380,6 +380,22 @@ def test_score(self):
                 score_return
             )
 
+    def test_teach(self):
+        X_training = np.random.rand(10, 2)
+        y_training = np.random.randint(0, 2, size=10)
+
+        for n_samples in range(1, 10):
+            X = np.random.rand(n_samples, 2)
+            y = np.random.randint(0, 2, size=n_samples)
+
+            learner = modAL.models.ActiveLearner(
+                X_training=X_training, y_training=y_training,
+                estimator=mock.MockClassifier()
+            )
+
+            learner.teach(X, y, only_new=False)
+            learner.teach(X, y, only_new=True)
+
     def test_keras(self):
         pass