Add note to training new models.

docs/training-new-models.md

@@ -1,6 +1,16 @@
-# Training new models
-This repository also contains our training infrastructure to promote an
+# Training new neural network models
+
+We have also released our training infrastructure to promote an
 open ecosystem and enable quicker bootstrapping for new research and development.
+
+The deep neural network (DNN) model produces the feature representation
+of faces.
+We provide pre-trained models trained with about 500k images that
+can be used to train new face classifiers as shown in
+[Demo 3](http://cmusatyalab.github.io/openface/demo-3-classifier/).
+Training a new DNN model is for advanced users and should be done
+with large datasets (>500k images) to improve the feature representation.
+
 Warning: Training is computationally and memory expensive and takes a
 few weeks on our Tesla K40 GPU.
 Because of this, the training code assumes CUDA is installed.