Add custom kernel cookbook

doc/cookbook/source/examples/kernel/custom_kernel.rst

@@ -0,0 +1,37 @@
+=============
+Custom Kernel
+=============
+
+The Custom Kernel allows for custom user provided kernel matrices.
+
+-------
+Example
+-------
+In this example, we initialize a :sgclass:`CGaussianKernel` and then use the kernel matrix to create a custom kernel.
+
+Imagine we have files with data.
+We create :sgclass:`CDenseFeatures` (here 64 bit floats aka RealFeatures) as
+
+.. sgexample:: custom_kernel_machine:create_features
+
+We create an instance of :sgclass:`CGaussianKernel`.
+We initialize with :sgclass:`CDenseFeatures` and then get the kernel matrix.
+
+.. sgexample:: custom_kernel_machine:create_kernel
+
+We use the provided kernel matrix to create custom kernels.
+
+.. sgexample:: custom_kernel_machine:create_custom_kernel
+
+We create an instance of :sgclass:`CLibSVM` with the custom kernel.
+
+.. sgexample:: custom_kernel_machine:create_machine
+
+Then we train the :sgclass:`CLibSVM` and we apply it to the test data, which gives the predicted :sgclass:`CBinaryLabels`.
+
+.. sgexample:: custom_kernel_machine:train_and_apply
+
+Finally, we can evaluate the performance, e.g. using :sgclass:`CAccuracyMeasure`.
+
+.. sgexample:: custom_kernel_machine:evaluate_accuracy
+

examples/meta/src/kernel/custom_kernel_machine.sg

@@ -16,9 +16,12 @@ k_obj.init(features_train, features_train)
 RealMatrix km_train = k_obj.get_kernel_matrix()
 k_obj.init(features_train, features_test)
 RealMatrix km_test = k_obj.get_kernel_matrix()
+#![create_kernel]
+
+#![create_custom_kernel]
 Kernel k_train = kernel(km_train)
 Kernel k_test = kernel(km_test)
-#![create_kernel]
+#![create_custom_kernel]
 
 #![create_machine]
 Machine svm = machine("LibSVM", C1=1.0, C2=1.0, kernel=k_train, labels=labels_train, epsilon=0.001)