scikit-learn · ogrisel · Mar 30, 2022 · Dec 7, 2020 · Dec 9, 2020 · Dec 9, 2020
diff --git a/doc/whats_new/v1.1.rst b/doc/whats_new/v1.1.rst
@@ -261,6 +261,24 @@ Changelog
   get accurate results when the number of features is large.
   :pr:`21109` by :user:`Smile <x-shadow-man>`.
 
+- |Enhancement| The :class:`decomposition.MiniBatchDictionaryLearning` and
+  :func:`decomposition.dict_learning_online` have been refactored and now have a
+  stopping criterion based on a small change of the dictionary or objective function,
+  controlled by the new `max_iter`, `tol` and `max_no_improvement` parameters. In
+  addition, some of their parameters and attributes are deprecated.
+
+  - the `n_iter` parameter of both is deprecated. Use `max_iter` instead.
+  - the `iter_offset`, `return_inner_stats`, `inner_stats` and `return_n_iter`
+    parameters of :func:`decomposition.dict_learning_online` serve internal purpose
+    and are deprecated.
+  - the `inner_stats_`, `iter_offset_` and `random_state_` attributes of
+    :class:`decomposition.MiniBatchDictionaryLearning` serve internal purpose and are
+    deprecated.
+  - the default value of the `batch_size` parameter of both will change from 3 to 256
+    in version 1.3.
+
+  :pr:`18975` by :user:`Jérémie du Boisberranger <jeremiedbb>`.
+
 - |Enhancement| :func:`decomposition.dict_learning`, :func:`decomposition.dict_learning_online`
   and :func:`decomposition.sparse_encode` preserve dtype for `numpy.float32`.
   :class:`decomposition.DictionaryLearning`, :class:`decompsition.MiniBatchDictionaryLearning`

diff --git a/examples/decomposition/plot_image_denoising.py b/examples/decomposition/plot_image_denoising.py
@@ -106,7 +106,7 @@ def show_with_diff(image, reference, title):
 data = data.reshape(data.shape[0], -1)
 data -= np.mean(data, axis=0)
 data /= np.std(data, axis=0)
-print("done in %.2fs." % (time() - t0))
+print(f"{data.shape[0]} patches extracted in %.2fs." % (time() - t0))
 
 
 # %%
@@ -116,10 +116,17 @@ def show_with_diff(image, reference, title):
 
 print("Learning the dictionary...")
 t0 = time()
-dico = MiniBatchDictionaryLearning(n_components=50, alpha=1, n_iter=250)
+dico = MiniBatchDictionaryLearning(
+    # increase to 300 for higher quality results at the cost of slower
+    # training times.
+    n_components=50,
+    batch_size=200,
+    alpha=1.0,
+    max_iter=10,
+)
 V = dico.fit(data).components_
 dt = time() - t0
-print("done in %.2fs." % dt)
+print(f"{dico.n_iter_} iterations / {dico.n_steps_} steps in {dt:.2f}.")
 
 plt.figure(figsize=(4.2, 4))
 for i, comp in enumerate(V[:100]):