120 use hinge loss for svm example (#124)

* Switch to hinge loss in SVM example

* remove named arguments in loss calls to allow the use of more losses

examples/plot_variable_importance_classif.py

@@ -29,7 +29,7 @@
 from scipy.stats import ttest_1samp
 from sklearn.base import clone
 from sklearn.linear_model import RidgeCV
-from sklearn.metrics import log_loss
+from sklearn.metrics import hinge_loss
 from sklearn.model_selection import RandomizedSearchCV, StratifiedKFold
 from sklearn.svm import SVC
 
@@ -163,7 +163,7 @@
         imputation_model=clone(imputation_model),
         n_permutations=50,
         n_jobs=5,
-        loss=log_loss,
+        loss=hinge_loss,
         random_state=seed,
         method="decision_function",
     )
@@ -177,7 +177,7 @@
         imputation_model=clone(imputation_model),
         n_permutations=50,
         n_jobs=5,
-        loss=log_loss,
+        loss=hinge_loss,
         random_state=seed,
         method="decision_function",
     )

src/hidimstat/cpi.py

@@ -236,7 +236,9 @@ def score(self, X, y):
 
         y_pred = getattr(self.estimator, self.method)(X)
 
-        loss_reference = self.loss(y_true=y, y_pred=y_pred)
+        # In sklearn API y_true is the first argument. Not specifying `y_true=...`
+        # will allows using other losses such as `hinge_loss`.
+        loss_reference = self.loss(y, y_pred)
         out_dict["loss_reference"] = loss_reference
 
         y_pred_perm = self.predict(X, y)
@@ -245,7 +247,7 @@ def score(self, X, y):
         for j, y_pred_j in enumerate(y_pred_perm):
             list_loss_perm = []
             for y_pred_perm in y_pred_j:
-                list_loss_perm.append(self.loss(y_true=y, y_pred=y_pred_perm))
+                list_loss_perm.append(self.loss(y, y_pred_perm))
             out_dict["loss_perm"][j] = np.array(list_loss_perm)
 
         out_dict["importance"] = np.array(

src/hidimstat/loco.py

@@ -128,7 +128,7 @@ def predict(self, X, y):
         output_dict = dict()
 
         y_pred = getattr(self.estimator, self.method)(X)
-        loss_reference = self.loss(y_true=y, y_pred=y_pred)
+        loss_reference = self.loss(y, y_pred)
         output_dict["loss_reference"] = loss_reference
         output_dict["loss_loco"] = dict()
 
@@ -183,13 +183,13 @@ def score(self, X, y):
         out_dict = dict()
         y_pred = getattr(self.estimator, self.method)(X)
 
-        loss_reference = self.loss(y_true=y, y_pred=y_pred)
+        loss_reference = self.loss(y, y_pred)
         out_dict["loss_reference"] = loss_reference
 
         y_pred_loco = self.predict(X, y)
 
         out_dict["loss_loco"] = np.array(
-            [self.loss(y_true=y, y_pred=y_pred_loco[j]) for j in range(self.n_groups)]
+            [self.loss(y, y_pred_loco[j]) for j in range(self.n_groups)]
         )
 
         out_dict["importance"] = out_dict["loss_loco"] - loss_reference

src/hidimstat/permutation_importance.py

@@ -172,7 +172,7 @@ def score(self, X, y):
 
         output_dict = dict()
         y_pred = getattr(self.estimator, self.method)(X)
-        loss_reference = self.loss(y_true=y, y_pred=y_pred)
+        loss_reference = self.loss(y, y_pred)
         output_dict["loss_reference"] = loss_reference
         output_dict["loss_perm"] = dict()
 
@@ -182,7 +182,7 @@ def score(self, X, y):
         for j, y_pred_j in enumerate(y_pred_perm):
             list_loss_perm = []
             for y_pred_perm in y_pred_j:
-                list_loss_perm.append(self.loss(y_true=y, y_pred=y_pred_perm))
+                list_loss_perm.append(self.loss(y, y_pred_perm))
             output_dict["loss_perm"][j] = np.array(list_loss_perm)
 
         output_dict["importance"] = np.array(