diff --git a/doc/modules/svm.rst b/doc/modules/svm.rst
index 0ac34cdcb6a10..8f97b8dee8806 100644
--- a/doc/modules/svm.rst
+++ b/doc/modules/svm.rst
@@ -521,9 +521,8 @@ is advised to use :class:`~sklearn.model_selection.GridSearchCV` with
  * :ref:`sphx_glr_auto_examples_svm_plot_rbf_parameters.py`
  * :ref:`sphx_glr_auto_examples_svm_plot_svm_nonlinear.py`
 
-|details-start|
-**Custom Kernels**
-|details-split|
+Custom Kernels
+--------------
 
 You can define your own kernels by either giving the kernel as a
 python function or by precomputing the Gram matrix.
@@ -539,8 +538,9 @@ classifiers, except that:
       use of ``fit()`` and ``predict()`` you will have unexpected results.
 
 
-Using Python functions as kernels
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+|details-start|
+**Using Python functions as kernels**
+|details-split|  
 
 You can use your own defined kernels by passing a function to the
 ``kernel`` parameter.
@@ -558,13 +558,13 @@ instance that will use that kernel::
     ...     return np.dot(X, Y.T)
     ...
     >>> clf = svm.SVC(kernel=my_kernel)
+  
+|details-end|
 
-.. topic:: Examples:
-
- * :ref:`sphx_glr_auto_examples_svm_plot_custom_kernel.py`.
 
-Using the Gram matrix
-~~~~~~~~~~~~~~~~~~~~~
+|details-start|
+**Using the Gram matrix**
+|details-split|  
 
 You can pass pre-computed kernels by using the ``kernel='precomputed'``
 option. You should then pass Gram matrix instead of X to the `fit` and
@@ -589,6 +589,10 @@ test vectors must be provided:
 
 |details-end|
 
+.. topic:: Examples:
+
+ * :ref:`sphx_glr_auto_examples_svm_plot_custom_kernel.py`.
+
 .. _svm_mathematical_formulation:
 
 Mathematical formulation