From 15456286da9378df6af796496ced5313ad7169f4 Mon Sep 17 00:00:00 2001
From: Jonathan L Long <jonlong@cs.berkeley.edu>
Date: Sat, 6 Sep 2014 21:39:56 -0700
Subject: [PATCH] [docs] tutorial/layers: brief descriptions of some loss
 layers

---
 docs/tutorial/layers.md | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/docs/tutorial/layers.md b/docs/tutorial/layers.md
index 11ca70e..036d9ad 100644
--- a/docs/tutorial/layers.md
+++ b/docs/tutorial/layers.md
@@ -123,17 +123,21 @@ Loss drives learning by comparing an output to a target and assigning cost to mi
 
 #### Softmax
 
-`SOFTMAX_LOSS`
+* LayerType: `SOFTMAX_LOSS`
+
+The softmax loss layer computes the multinomial logistic loss of the softmax of its inputs. It's conceptually identical to a softmax layer followed by a multinomial logistic loss layer, but provides a more numerically stable gradient.
 
 #### Sum-of-Squares / Euclidean
 
-`EUCLIDEAN_LOSS`
+* LayerType: `EUCLIDEAN_LOSS`
+
+The Euclidean loss layer computes the sum of squares of differences of its two inputs, $$\frac 1 {2N} \sum_{i=1}^N \| x^1_i - x^2_i \|_2^2$$.
 
 #### Hinge / Margin
 
 * LayerType: `HINGE_LOSS`
 * CPU implementation: `./src/caffe/layers/hinge_loss_layer.cpp`
-* CUDA GPU implementation: `NOT_AVAILABLE`
+* CUDA GPU implementation: none yet
 * Parameters (`HingeLossParameter hinge_loss_param`)
     - Optional
         - `norm` [default L1]: the norm used. Currently L1, L2
@@ -164,6 +168,8 @@ Loss drives learning by comparing an output to a target and assigning cost to mi
         }
       }
 
+The hinge loss layer computes a one-vs-all hinge or squared hinge loss.
+
 #### Sigmoid Cross-Entropy
 
 `SIGMOID_CROSS_ENTROPY_LOSS`
-- 
2.7.4