Changed the stopping condition for the structured svm to the one suggested

by the Joachims paper.  I also added another useful reference to the docs.

--HG--
extra : convert_revision : svn%3Afdd8eb12-d10e-0410-9acb-85c331704f74/trunk%404247

dlib/svm/structural_svm_problem.h

@@ -164,9 +164,7 @@ namespace dlib
 
             bool should_stop = false;
 
-            if (current_objective_value == 0)
-                should_stop = true;
-            else if (current_error_gap/current_objective_value < eps)
+            if (current_risk_gap < eps)
                 should_stop = true;
 
             if (should_stop && !skip_cache)

dlib/svm/structural_svm_problem_abstract.h

@@ -65,14 +65,20 @@ namespace dlib
                     and C > 0
 
                 
-                For further information you should consult the following paper: 
+
+                For an introduction to structured support vector machines you should consult 
+                the following paper: 
+                    Predicting Structured Objects with Support Vector Machines by 
+                    By Thorsten Joachims, Thomas Hofmann, Yisong Yue, and Chun-nam Yu
+
+                For a more detailed discussion of the particular algorithm implemented by this
+                object see the following paper:  
                     T. Joachims, T. Finley, Chun-Nam Yu, Cutting-Plane Training of Structural SVMs, 
                     Machine Learning, 77(1):27-59, 2009.
 
                     Note that this object is essentially a tool for solving the 1-Slack structural
                     SVM with margin-rescaling.  Specifically, see Algorithm 3 in the above referenced 
                     paper.
-
         !*/
 
         typedef typename matrix_type::type scalar_type;
@@ -101,7 +107,8 @@ namespace dlib
             ensures
                 - returns the error epsilon that determines when training should stop.
                   Smaller values may result in a more accurate solution but take longer 
-                  to execute.
+                  to execute.  Specifically, the algorithm stops when the average sample
+                  risk (i.e. R(w) as defined above) is within epsilon of its optimal value.
         !*/
 
         void set_max_cache_size (