Adding clarifications to spec which were pointed out by Scott Richardson.

dlib/dnn/core_abstract.h

@@ -242,7 +242,7 @@ namespace dlib
         );
         /*!
             ensures
-                - #layer_details() == layer_details_type(layer_det)
+                - #layer_details() == layer_det
                 - #subnet()        == subnet_type(args)
         !*/
 
@@ -562,7 +562,7 @@ namespace dlib
         ); 
         /*!
             ensures
-                - #loss_details() == loss_details_type(layer_det)
+                - #loss_details() == layer_det
                 - #subnet()       == subnet_type(args)
         !*/
 
@@ -573,7 +573,7 @@ namespace dlib
         );
         /*!
             ensures
-                - #loss_details() == loss_details_type(layer_det)
+                - #loss_details() == layer_det
                 - #subnet()       == subnet_type(args)
         !*/
 

dlib/dnn/layers_abstract.h

@@ -30,7 +30,11 @@ namespace dlib
                                       \__ subnetwork for layer1 __/
 
                 Therefore, by "subnetwork" we mean the part of the network closer to the
-                input.  
+                input.
+
+                Note that there is no dlib::SUBNET type.  It is shown here purely to
+                document the interface layer objects expect to see when they interact
+                with a network.
         !*/
 
     public:
@@ -207,7 +211,7 @@ namespace dlib
                   Moreover, this was the most recent call to forward().  This means that
                   forward() is allowed to cache intermediate results so they can be used
                   during the backward computation.
-                - have_same_dimensions(gradient_input, computed_output)
+                - have_same_dimensions(gradient_input, computed_output) == true
                 - have_same_dimensions(sub.get_gradient_input(), sub.get_output()) == true
                 - have_same_dimensions(params_grad, get_layer_params()) == true
             ensures

dlib/dnn/tensor_abstract.h

@@ -446,7 +446,7 @@ namespace dlib
         !*/
     };
 
-    void serialize(const tensor& item, std::ostream& out)
+    void serialize(const tensor& item, std::ostream& out);
     void deserialize(resizable_tensor& item, std::istream& in);
     /*!
         provides serialization support for tensor and resizable_tensor.  Note that you can
@@ -553,7 +553,7 @@ namespace dlib
         !*/
     };
 
-    void serialize(const alias_tensor& item, std::ostream& out)
+    void serialize(const alias_tensor& item, std::ostream& out);
     void deserialize(alias_tensor& item, std::istream& in);
     /*!
         provides serialization support for alias_tensor.