Add docs new method + change score base

Author: antoinedemathelin

adapt/base.py

@@ -178,7 +178,7 @@ def pretrain_step(self, data):
             ys_pred = tf.reshape(ys_pred, tf.shape(ys))
 
             # Compute the loss value
-            loss = self.task_loss_(ys, ys_pred)
+            loss = tf.reduce_mean(self.task_loss_(ys, ys_pred))
             task_loss = loss + sum(self.task_.losses)
             enc_loss = loss + sum(self.encoder_src_.losses)
 

adapt/feature_based/_adda.py

@@ -185,7 +185,7 @@ def transform(self, X, domain="tgt"):
             Input data.
 
         domain : str (default="tgt")
-            Choose between ``"source", "src"`` and
+            Choose between ``"source", "src"`` or
             ``"target", "tgt"`` feature embedding.
 
         Returns

adapt/feature_based/_coral.py

@@ -92,7 +92,7 @@ class fMMD(BaseAdaptEstimator):
     """
     fMMD : feature Selection with MMD
     
-    LDM selects input features inorder to minimize the
+    LDM selects input features in order to minimize the
     maximum mean discrepancy (MMD) between the source and
     the target data.
     
@@ -160,7 +160,25 @@ def __init__(self,
         super().__init__(**kwargs)
 
 
-    def fit_transform(self, Xs, Xt, **fit_params):
+    def fit_transform(self, Xs, Xt, **kwargs):
+        """
+        Fit embeddings.
+        
+        Parameters
+        ----------
+        Xs : array
+            Input source data.
+            
+        Xt : array
+            Input target data.
+                        
+        kwargs : key, value argument
+            Not used, present here for adapt consistency.
+            
+        Returns
+        -------
+        Xs_emb : embedded source data
+        """
         Xs = check_array(Xs)
         Xt = check_array(Xt)
         set_random_seed(self.random_state)
@@ -217,5 +235,18 @@ def F(x=None, z=None):
 
 
     def transform(self, X):
+        """
+        Return the projection of X on the selected featues.
+
+        Parameters
+        ----------
+        X : array
+            Input data.
+
+        Returns
+        -------
+        X_emb : array
+            Embeddings of X.
+        """
         X = check_array(X)
         return X[:, self.selected_features_]

adapt/feature_based/_fmmd.py

@@ -223,6 +223,27 @@ def train_step(self, data):
             return logs
 
 
+    def predict_avg(self, X):
+        """
+        Return the average predictions between
+        task_ and discriminator_ networks.
+        
+        Parameters
+        ----------
+        X : array
+            Input data
+            
+        Returns
+        -------
+        y_avg : array
+            Average predictions
+        """
+        ypt = self.task_.predict(self.transform(X))
+        ypd = self.discriminator_.predict(self.transform(X))
+        yp_avg = 0.5 * (ypt+ypd)
+        return yp_avg
+        
+
     def _initialize_networks(self):
         if self.encoder is None:
             self.encoder_ = get_default_encoder(name="encoder")

adapt/feature_based/_mcd.py

@@ -49,7 +49,25 @@ def __init__(self,
         super().__init__(**kwargs)
 
 
-    def fit_transform(self, Xs, Xt, **fit_params):
+    def fit_transform(self, Xs, Xt, **kwargs):
+        """
+        Fit embeddings.
+        
+        Parameters
+        ----------
+        Xs : array
+            Input source data.
+            
+        Xt : array
+            Input target data.
+                        
+        kwargs : key, value argument
+            Not used, present here for adapt consistency.
+            
+        Returns
+        -------
+        Xs_emb : embedded source data
+        """
         Xs = check_array(Xs)
         Xt = check_array(Xt)
         set_random_seed(self.random_state)
@@ -67,6 +85,29 @@ def fit_transform(self, Xs, Xt, **fit_params):
 
 
     def transform(self, X, domain="tgt"):
+        """
+        Project X in the target subspace.
+        
+        The paramter ``domain`` specify if X should
+        be considered as source or target data. As the
+        transformation is assymetric, the source transformation
+        should be applied on source data and the target
+        transformation on target data.
+
+        Parameters
+        ----------
+        X : array
+            Input data.
+            
+        domain : str (default="tgt")
+            Choose between ``"source", "src"`` or
+            ``"target", "tgt"`` feature embedding.
+
+        Returns
+        -------
+        X_emb : array
+            Embeddings of X.
+        """
         X = check_array(X)
 
         if domain in ["tgt", "target"]:

adapt/feature_based/_sa.py

@@ -66,6 +66,24 @@ def __init__(self,
 
 
     def fit_weights(self, Xs, Xt, **kwargs):
+        """
+        Fit importance weighting.
+        
+        Parameters
+        ----------
+        Xs : array
+            Input source data.
+            
+        Xt : array
+            Input target data.
+            
+        kwargs : key, value argument
+            Not used, present here for adapt consistency.
+            
+        Returns
+        -------
+        weights_ : sample weights
+        """
         Xs = check_array(Xs)
         Xt = check_array(Xt)
         set_random_seed(self.random_state)

adapt/instance_based/_ldm.py

@@ -100,6 +100,24 @@ def __init__(self,
 
 
     def fit_weights(self, Xs, Xt, **kwargs):
+        """
+        Fit importance weighting.
+        
+        Parameters
+        ----------
+        Xs : array
+            Input source data.
+            
+        Xt : array
+            Input target data.
+            
+        kwargs : key, value argument
+            Not used, present here for adapt consistency.
+            
+        Returns
+        -------
+        weights_ : sample weights
+        """
         Xs = check_array(Xs)
         Xt = check_array(Xt)
         set_random_seed(self.random_state)