Added doc for itml

Author: terrytangyuan

doc/source/algorithms/itml.rst

@@ -1,8 +1,19 @@
 Information Theoretic Metric Learning (ITML)
 =====================================
+ITML minimizes the differential relative entropy between two multivariate Gaussians under constraints on the distance function, which can be formulated into a particular Bregman optimization problem by minimizing the LogDet divergence subject to the linear constraints. This algorithm can handle wide variety of constraints and can optionally incorporate a prior on the distance function. Unlike some existing method, ITML is fast and scalable since no eigenvalue computation or semi-definite programming are required. 
 
 Example Code
 ------------------
+After loading the iris data, we apply ITML:
+::
+	from metric_learn import ITML
+
+	num_constraints = 200
+	n = self.iris_points.shape[0]
+	C = ITML.prepare_constraints(self.iris_labels, n, num_constraints)
+	itml = ITML().fit(self.iris_points, C, verbose=False)
+	itml.transform()
 
 References
-------------------
+------------------
+`Information-theoretic Metric Learning <http://machinelearning.wustl.edu/mlpapers/paper_files/icml2007_DavisKJSD07.pdf>`_ Davis, Jason V., et al.

doc/source/algorithms/lmnn.rst

@@ -8,6 +8,7 @@ Example Code
 We use iris data here for all the examples:
 ::
 	from sklearn.datasets import load_iris
+	import numpy as np
 
 	iris_data = load_iris()
 	self.iris_points = iris_data['data']
@@ -16,12 +17,13 @@ We use iris data here for all the examples:
 
 In this package, we have two different implementations of LMNN. Here we try both implementations in a for loop:
 ::
+	from metric_learn import LMNN
+	from metric_learn.lmnn import python_LMNN
+
 	for LMNN_cls in set((LMNN, python_LMNN)):
 	      lmnn = LMNN_cls(k=k, learn_rate=1e-6)
 	      lmnn.fit(self.iris_points, self.iris_labels, verbose=False)
-
-	      csep = class_separation(lmnn.transform(), self.iris_labels)
-	      self.assertLess(csep, 0.25)
+	      lmnn.transform()
 
 References
 ------------------