Adding example doxcumentation

contrib/Python/doc/examples.rst

@@ -1,3 +1,27 @@
 Examples
 ========
 
+Logistic Regression
+-------------------
+
+Examples for logistic regression you find here: `<https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LogReg/>`_ 
+
+- Using training and testing data *from a file* : `logreg.py <https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LogReg/logreg.py>`_ .
+- Using training and testing data *from a NumPy array* : `logreg_numpy.py <https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LogReg/logreg_numpy.py>`_ .
+
+LSTM-based sequence classification
+----------------------------------
+Example for training an LSTM-based sequence classification model with embedding you find here: `<https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LSTM/>`_ .
+A typical application would be e.g. a text classification where we leverage a precomputed word-embedding. 
+This is a good example to see how to provide input data for sequences and using sparse input.
+
+- In  `Train_sparse.txt <https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LSTM/>`_  we have two inputs. The input *x* provides the sequence data in sparse form, while *y* provides the classes in dense form.
+- The example also uses a predefined embedding (`embeddingmatrix.txt <https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/LSTM/embeddingmatrix.txt>`_ ) mapping each index of the input sequence *x* to an embedding vector.
+
+One hidden layer neural network
+-------------------------------
+ 
+Example for training a *one hidden layer neural network* using the MNIST-data (recognition of handwritten digits) you find here: `<https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/MNIST/>`_ .
+
+To obtain and prepare the MNIST data use `fetch_mnist_data.py <https://github.com/Microsoft/CNTK/tree/master/contrib/Python/cntk/examples/MNIST/fetch_mnist_data.py>`_ .
+