[FIX][ENH] Improve "DA methods" examples (#188)

* different section for pipeline and adapter * same for mapping methods * typo * update how to use skada * oversight * update how to use * update README and users guide * Update "DA methods" examples * update * update --------- Co-authored-by: apmellot <[email protected]> Co-authored-by: Rémi Flamary <[email protected]> Co-authored-by: Théo Gnassounou <[email protected]> Co-authored-by: tgnassou <[email protected]>
scikit-adaptation · Jul 4, 2024 · 0d9c6b1 · 0d9c6b1
1 parent e2db517
commit 0d9c6b1
Show file tree

Hide file tree

Showing 4 changed files with 23 additions and 19 deletions.
diff --git a/examples/methods/plot_dasvm.py b/examples/methods/plot_dasvm.py
@@ -39,8 +39,8 @@
 # We generate our 2D dataset with 2 classes
 # ------------------------------------------
 #
-# We generate a simple 2D dataset from moon distribution, where source and target
-# are not taken from the same location in the moons. This dataset thus present
+# We generate a simple 2D dataset from a moon distribution, where source and target
+# are not taken from the same location in the moons. This dataset thus presents a
 # covariate shift.
 
 X, y, sample_domain = make_dataset_from_moons_distribution(
@@ -86,7 +86,7 @@
 # The main problem here is that we only know the distribution of the points
 # from the target dataset, our goal is to label it.
 #
-# The algorithm of the DASVM consist in fitting multiple base_estimator (SVC) by:
+# The DASVM method consist in fitting multiple base_estimator (SVC) by:
 #     - Removing from the training dataset (if possible)
 #       `k` points from the source dataset for which the current
 #       estimator is doing well

diff --git a/examples/methods/plot_reweighting.py b/examples/methods/plot_reweighting.py
@@ -35,12 +35,12 @@
 
 # %%
 #
-#     The reweighting methods
+# Reweighting Methods
 # ------------------------------------------
-# The goal of reweighting methods is to estimate some weights for the
-# source dataset in order to then fit a estimator on the source dataset,
-# while taking those weights into account, so that the fitted estimator is
-# well suited to predicting labels from points drawn from the target distribution.
+# The purpose of reweighting methods is to estimate weights for the source dataset.
+# These weights are then used to fit an estimator on the source dataset, taking the
+# weights into account. The goal is to ensure that the fitted estimator is suitable
+# for predicting labels from the target distribution.
 #
 # Reweighting methods implemented and illustrated are the following:
 #   * :ref:`Density Reweighting<Illustration of the Density Reweighting method>`

diff --git a/examples/methods/plot_subspace.py b/examples/methods/plot_subspace.py
@@ -33,14 +33,15 @@
 #     The subspaces methods
 # ------------------------------------------
 #
-# Firstly we are in the case of unsupervised domain adaptation, it means that
-# we know the labels of the source data but not from the target data.
-# The goal of subspace is to project data from a d dimensional space
-# into a d' dimensional space with d'<d.
-# This kind of da method is especially good when we work with subspace
-# shift, meaning that there is a subspace on which, when projected on it,
-# the source and target data have the same distributions.
-#
+# Supspace methods are used in unsupervised domain adaptation.
+# In this case, we have labeled data for the source domain but not for the target
+# domain.
+# The goal of subspace methods is to project data from a d-dimensional space
+# into a lower-dimensional space with d' < d.
+# Subspace methods are particularly effective when dealing with subspace shift,
+# where the source and target data have the same distributions when projected onto a
+# subspace.
+
 # The Subspace methods implemented and illustrated are the following:
 #   * :ref:`Subspace Alignment<Illustration of the Subspace Alignment method>`
 #   * :ref:`Transfer Component Analysis<Illustration of the Transfer Component
@@ -339,7 +340,7 @@ def plot_subspace_and_classifier(
 #
 # In most of the previous works, we explored two learning strategies independently for
 # domain adaptation: feature matching and instance reweighting. Transfer Joint Matching
-# (TJM) aims to use both, by adding a constant to tradeoff between to two.
+# (TJM) aims to use both, by adding a constant to tradeoff between the two.
 #
 # See [26] for details:
 #

diff --git a/examples/plot_how_to_use_skada.py b/examples/plot_how_to_use_skada.py
@@ -97,9 +97,10 @@
 # DA estimator in a pipeline
 # -----------------------------
 #
+
 # SKADA estimators can be used as the final estimator of a scikit-learn pipeline.
-# Again, the only difference is that the :code:`sample_domain` array must be
-# passed by name during in fit.
+# Again, the only difference is that the :code:`sample_domain` array must be passed
+# by name during in fit.
 
 
 # create a DA pipeline
@@ -120,6 +121,7 @@
 # Here is an example with the CORAL and GaussianReweight adapters.
 #
 # .. WARNING::
+
 #   Note that as illustrated below for reweighting adapters, one needs a
 #   subsequent estimator that takes :code:`sample_weight` as an input parameter.
 #   This can be done using the :code:`set_fit_request` method of the estimator
@@ -222,6 +224,7 @@
 print("Accuracy on source:", pipe.score(Xs, ys, sample_domain=sample_domain_s))
 print("Accuracy on target:", pipe.score(Xt, yt))  # target by default
 
+
 # %%
 # Similarly one can use the PerDomain selector to train a different estimator
 # per domain. This allows to handle multiple source and target domains. In this