nn4.def: Add reasoning behind LRN layers.

models/openface/nn4.def.lua

@@ -28,6 +28,17 @@ function createModel()
    net:add(nn.SpatialBatchNormalization(64))
    net:add(nn.ReLU())
 
+   -- The FaceNet paper just says `norm` and that the models are based
+   -- heavily on the inception paper (http://arxiv.org/pdf/1409.4842.pdf),
+   -- which uses pooling and normalization in the same way in the early layers.
+   --
+   -- The Caffe and official versions of this network both use LRN:
+   --
+   --   + https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet
+   --   + https://github.com/google/inception/blob/master/inception.ipynb
+   --
+   -- The Caffe docs at http://caffe.berkeleyvision.org/tutorial/layers.html
+   -- define LRN to be across channels.
    net:add(nn.SpatialMaxPooling(3, 3, 2, 2, 1, 1))
    net:add(nn.CrossMapNormalization(5, 0.0001, 0.75))