Fixed some warnings and errors from visual studio 2015

dlib/dnn/core.h

@@ -876,7 +876,7 @@ namespace dlib
             "The loss layer and input layer must agree on the sample_expansion_factor.");
 
 
-        add_loss_layer() = default;
+        add_loss_layer() {};
         add_loss_layer(const add_loss_layer&) = default;
         add_loss_layer(add_loss_layer&&) = default;
         add_loss_layer& operator=(add_loss_layer&&) = default;
@@ -1478,7 +1478,7 @@ namespace dlib
 
         // ==================================================================
         // first validate the way the parameter gradients are computed
-        for (long i = 0; i < params_grad.size(); ++i)
+        for (unsigned long i = 0; i < params_grad.size(); ++i)
         {
             layer_details_type l1(l);
 

dlib/dnn/loss.h

@@ -35,7 +35,7 @@ namespace dlib
             DLIB_CASSERT(output_tensor.num_samples()%sample_expansion_factor == 0,"");
 
             const float* out_data = output_tensor.host();
-            for (unsigned long i = 0; i < output_tensor.num_samples(); ++i)
+            for (long i = 0; i < output_tensor.num_samples(); ++i)
             {
                 *iter++ = out_data[i];
             }
@@ -67,7 +67,7 @@ namespace dlib
             double loss = 0;
             const float* out_data = output_tensor.host();
             float* g = grad.host();
-            for (unsigned long i = 0; i < output_tensor.num_samples(); ++i)
+            for (long i = 0; i < output_tensor.num_samples(); ++i)
             {
                 const float y = *truth++;
                 DLIB_CASSERT(y == +1 || y == -1, "y: " << y);

dlib/dnn/trainer.h

@@ -123,7 +123,7 @@ namespace dlib
             {
                 running_stats<double> rs;
 
-                unsigned long j = 0;
+                size_t j = 0;
 
                 // Load two tensors worth of data at once so we can overlap the computation
                 // and data transfer between the host and the device.
@@ -140,7 +140,7 @@ namespace dlib
                     j += mini_batch_size;
                 }
 
-                unsigned long i = 0;
+                size_t i = 0;
                 using namespace std::chrono;
                 auto last_time = system_clock::now();
                 while (i < data.size())
@@ -211,7 +211,7 @@ namespace dlib
             for (unsigned long epoch_iteration = 0; epoch_iteration < num_epochs; ++epoch_iteration)
             {
                 running_stats<double> rs;
-                unsigned long j = 0;
+                size_t j = 0;
 
                 // Load two tensors worth of data at once so we can overlap the computation
                 // and data transfer between the host and the device.
@@ -228,7 +228,7 @@ namespace dlib
                     j += mini_batch_size;
                 }
 
-                unsigned long i = 0;
+                size_t i = 0;
                 using namespace std::chrono;
                 auto last_time = system_clock::now();
                 while (i < data.size())
@@ -318,7 +318,7 @@ namespace dlib
         }
 
         unsigned long num_epochs;
-        unsigned long mini_batch_size;
+        size_t mini_batch_size;
         bool verbose;
 
         net_type net;