Added the forces_last_weight_to_1() option to the svm_c_linear_trainer.

dlib/svm/svm_c_linear_trainer.h

@@ -352,6 +352,7 @@ namespace dlib
             eps = 0.001;
             max_iterations = 10000;
             learn_nonnegative_weights = false;
+            last_weight_1 = false;
         }
 
         explicit svm_c_linear_trainer (
@@ -372,6 +373,7 @@ namespace dlib
             eps = 0.001;
             max_iterations = 10000;
             learn_nonnegative_weights = false;
+            last_weight_1 = false;
         }
 
         void set_epsilon (
@@ -443,6 +445,19 @@ namespace dlib
             learn_nonnegative_weights = value;
         }
 
+        bool forces_last_weight_to_1 (
+        ) const
+        {
+            return last_weight_1;
+        }
+
+        void force_last_weight_to_1 (
+            bool should_last_weight_be_1
+        )
+        {
+            last_weight_1 = should_last_weight_be_1;
+        }
+
         void set_c (
             scalar_type C 
         )
@@ -555,16 +570,26 @@ namespace dlib
             typedef matrix<scalar_type,0,1> w_type;
             w_type w;
 
+            const unsigned long num_dims = max_index_plus_one(x);
+
             unsigned long num_nonnegative = 0;
             if (learn_nonnegative_weights)
             {
-                num_nonnegative = max_index_plus_one(x);
+                num_nonnegative = num_dims;
+            }
+
+            unsigned long force_weight_1_idx = std::numeric_limits<unsigned long>::max(); 
+            if (last_weight_1)
+            {
+                force_weight_1_idx = num_dims-1; 
             }
 
+
             svm_objective = solver(
                 make_oca_problem_c_svm<w_type>(Cpos, Cneg, x, y, verbose, eps, max_iterations), 
                 w,
-                num_nonnegative);
+                num_nonnegative,
+                force_weight_1_idx);
 
             // put the solution into a decision function and then return it
             decision_function<kernel_type> df;
@@ -589,6 +614,7 @@ namespace dlib
         bool verbose;
         unsigned long max_iterations;
         bool learn_nonnegative_weights;
+        bool last_weight_1;
     }; 
 
 // ----------------------------------------------------------------------------------------

dlib/svm/svm_c_linear_trainer_abstract.h

@@ -53,6 +53,7 @@ namespace dlib
                 - this object will not be verbose unless be_verbose() is called
                 - #get_max_iterations() == 10000
                 - #learns_nonnegative_weights() == false
+                - #force_last_weight_to_1() == false
         !*/
 
         explicit svm_c_linear_trainer (
@@ -71,6 +72,7 @@ namespace dlib
                 - this object will not be verbose unless be_verbose() is called
                 - #get_max_iterations() == 10000
                 - #learns_nonnegative_weights() == false
+                - #force_last_weight_to_1() == false
         !*/
 
         void set_epsilon (
@@ -170,6 +172,25 @@ namespace dlib
                 - #learns_nonnegative_weights() == value
         !*/
 
+        bool forces_last_weight_to_1 (
+        ) const;
+        /*!
+            ensures
+                - returns true if this trainer has the constraint that the last weight in
+                  the learned parameter vector must be 1.  This is the weight corresponding
+                  to the feature in the training vectors with the highest dimension.  
+                - Forcing the last weight to 1 also disables the bias and therefore the b
+                  field of the learned decision_function will be 0 when forces_last_weight_to_1() == true.
+        !*/
+
+        void force_last_weight_to_1 (
+            bool should_last_weight_be_1
+        );
+        /*!
+            ensures
+                - #forces_last_weight_to_1() == should_last_weight_be_1
+        !*/
+
         void set_c (
             scalar_type C 
         );