Fix segmentation fault when SVM is trained in sequenece. Fixes issue #9

This one is tricky. Although the tests are passing now (most of it), the solution is not appropriate.

I guess there is some reference counter that is decremented by mistake, but I was not able to find it. @andre.anjos look at the diff, I just created a fake iterator in the end of the training.

bob/learn/libsvm/test_trainer.py

@@ -214,4 +214,21 @@ def test_training_one_class():
   curr_scores = numpy.array(curr_scores)
   prev_scores = numpy.array(prev_scores)
   assert numpy.all(abs(curr_scores - prev_scores) < 1e-8)
- 
+
+
+def test_successive_training():
+
+  # Tests successive training works: i.e., training a couple of machines one
+  # after the other.
+
+  numpy.random.seed(10)
+
+  for i in range(2):
+    pos = numpy.random.normal(0., 1, size=(100, 2))
+    neg = numpy.random.normal(1., 1, size=(100, 2))
+    data = [pos, neg]
+
+    trainer = Trainer()
+    trainer.kernel_type = 'LINEAR'
+    trainer.cost = 1
+    trainer.train(data)