Now considering n_stable based on the training set if there's no vali…

…dation set specified. Updates and fixes to documentation and tests accrodingly.

README.rst

@@ -33,12 +33,12 @@ If you want to use the latest official release, you can do so from PYPI directly
 
     > pip install scikit-neuralnetwork
 
-This will install a copy of `Lasagne` too as a dependency. We recommend you use a virtual environment for Python.
+This will install a copy of ``Lasagne`` and other minor packages too as a dependency. We highly recommend you use a virtual environment for Python.
 
 B) Pulling Repositories [Optional]
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-You'll need to first install some dependencies manually
+You'll need some dependencies, which you can install manually as follows::
 
     > pip install numpy scipy theano lasagne
 

sknn/mlp.py

@@ -119,8 +119,8 @@ def _reshape(self, X, y=None):
         return X, y
 
     def _train(self, X, y):
-        assert self.n_iter or self.valid_set,\
-            "Neither n_iter nor valid_set were specified; training would loop forever."
+        assert self.n_iter or self.n_stable,\
+            "Neither n_iter nor n_stable were specified; training would loop forever."
 
         best_train_error, best_valid_error = float("inf"), float("inf")
         best_params = [] 
@@ -159,7 +159,7 @@ def _train(self, X, y):
                       time.time() - start
                       ))
 
-            if best_valid:
+            if best_valid or (self.valid_set is None and best_train):
                 best_params = self._backend._mlp_to_array()
                 n_stable = 0
             else:

sknn/nn.py

@@ -293,12 +293,16 @@ class NeuralNetwork(object):
 
     n_stable: int, optional
         Number of interations after which training should return when the validation
-        error remains constant.  This is a sign that the data has been fitted, or that
-        optimization may have stalled. Default is ``10``.
+        error remains (near) constant.  This is usually a sign that the data has been
+        fitted, or that optimization may have stalled.  If no validation set is specified,
+        then stability is judged based on the training error.  Default is ``10``.
 
     f_stable: float, optional
         Threshold under which the validation error change is assumed to be stable, to
-        be used in combination with `n_stable`. Default is ``0.001`.
+        be used in combination with `n_stable`. This is calculated as a relative ratio
+        of improvement, so if the results are only 0.1% better training is considered
+        stable. The training set is used as fallback if there's no validation set. Default
+        is ``0.001`.
 
     valid_set: tuple of array-like, optional
         Validation set (X_v, y_v) to be used explicitly while training.  Both

sknn/tests/test_training.py

@@ -36,7 +36,7 @@ def test_FitAutomaticValidation(self):
 
     def test_TrainingInfinite(self):
         a_in, a_out = numpy.zeros((8,16)), numpy.zeros((8,4))
-        self.nn = MLP(layers=[L("Linear")])
+        self.nn = MLP(layers=[L("Linear")], n_iter=None, n_stable=None)
         assert_raises(AssertionError, self.nn._fit, a_in, a_out)