Pairwise pairs

Added parallel loop for pairwise pairs

examples/plot_basic_usage.py

@@ -63,6 +63,9 @@ def score_function(X):
 
 print("Pairwise comparison (one vs group):", xai.pairwise(X[2], X[5:10]))
 
+pairlist=[(X[2], X[3]), (X[2], X[4]), (X[2], X[2]), (X[4], X[2])]
+print("Pairwise comparison (group of pairs):", xai.pairwise_all(pairlist))
+
 
 ######################################################################################
 # We can also turn these into visualizations:

sharp/base.py

@@ -258,3 +258,24 @@ def pairwise(self, sample1, sample2, **kwargs):
             coalition_size=coalition_size,
             **kwargs
         )
+
+    def pairwise_all(self, pairs, **kwargs):
+        """
+        set_cols_idx should be passed in kwargs if measure is marginal
+        pairs is a list of tuples of indexes
+        """
+        # X_ref = self._X if self._X is not None else check_inputs(X)[0]
+
+        if "sample_size" in kwargs.keys():
+            sample_size = 1
+
+        influences = parallel_loop(
+            lambda idx: self.individual(
+                pairs[idx][0].reshape(1, -1), X=pairs[idx][1].reshape(1, -1), verbose=False, **kwargs
+            ),
+            range(len(pairs)),
+            n_jobs=self.n_jobs,
+            progress_bar=self.verbose,
+        )
+
+        return np.array(influences)