>>> a=sklearn.svm.SVC(kernel=lambda x, y: None)
>>> a.kernel_function
<function <lambda> at 0xa769aac>
>>> sklearn.base.clone(a).kernel_function
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: 'SVC' object has no attribute 'kernel_function'
This makes it so that cross_validation_score can't operate on estimators with custom kernels.