scikit-learn · shanglun · Aug 7, 2016 · Aug 7, 2016 · Aug 7, 2016 · Aug 7, 2016
diff --git a/doc/developers/contributing.rst b/doc/developers/contributing.rst
@@ -798,6 +798,8 @@ The reason for postponing the validation is that the same validation
 would have to be performed in ``set_params``,
 which is used in algorithms like ``GridSearchCV``.
 
+.. _fitting:
+
 Fitting
 ^^^^^^^
 
@@ -974,6 +976,23 @@ The easiest and recommended way to accomplish this is to
 **not do any parameter validation in** ``__init__``.
 All logic behind estimator parameters,
 like translating string arguments into functions, should be done in ``fit``.
+While it is possible to pass parameters into the ``fit`` method, these
+should be restricted to variables that need to be sliced during
+cross-validation. These variables need to be arrays with ``shape[0]==n_samples``
+(see :ref:`fitting` above). An example of this type of variable would be
+``sample_weight``. This interface allows all data-dependent parameters to be
+sliced cleanly during cross-validation, and allows
+operations such as gridsearch and pipelining.
+All other parameters should be passed to
+``__init__`` or set after construction using ``set_params``.
+
+As a general rule, ``transform`` and ``predict`` should not take additional
+parameters. In cases where additional parameters would be useful, e.g. when
+setting a threshold value for feature selection, the custom parameters
+should be passed to ``__init__``. In case parameters need to be changed after
+the call to __init__, e.g. when setting a threshold value for feature
+selection, this can be done by setting the attribute directly on the estimator
+ or calling set_parms.
 
 Also it is expected that parameters with trailing ``_`` are **not to be set
 inside the** ``__init__`` **method**. All and only the public attributes set by