Added recommend_batch()

rtrec/models/base.py

@@ -132,6 +132,38 @@ def _recommend(self, user_id: int, candidate_item_ids: List[int], top_k: int = 1
         """
         raise NotImplementedError("_recommend method must be implemented in the derived class")
 
+    def recommend_batch(self, users: List[Any], top_k: int = 10, filter_interacted: bool = True) -> List[List[Any]]:
+        """
+        Recommend top-K items for a list of users.
+        :param users: List of user indices
+        :param top_k: Number of top items to recommend
+        :param filter_interacted: Whether to filter out items the user has already interacted with
+        :return: List of top-K item indices recommended for each user
+        """
+        user_ids = [self.user_ids.get_id(user) for user in users]
+        interaction_matrix = self.interactions.to_csr(select_users=user_ids)
+
+        results = []
+        for user_id in user_ids:
+            if user_id is None:
+                results.append([]) # TODO: return popoular items?
+                continue
+            recommended_item_ids = self._recommend_batch(user_id, interaction_matrix, top_k=top_k, filter_interacted=filter_interacted)
+            results.append([self.item_ids.get(item_id) for item_id in recommended_item_ids])
+        return results
+
+    def _recommend_batch(self, user_id: int, interaction_matrix: csc_matrix, candidate_item_ids: Optional[List[int]]=None, top_k: int = 10, filter_interacted: bool = True) -> List[int]:
+        """
+        Recommend top-K items for a list of users.
+        :param user_id: User index
+        :param interaction_matrix: Sparse user-item interaction matrix
+        :param candidate_item_ids: List of candidate item indices
+        :param top_k: Number of top items to recommend
+        :param filter_interacted: Whether to filter out items the user has already interacted with
+        :return: List of top-K item indices recommended for each user
+        """
+        raise NotImplementedError("_recommend_batch method must be implemented in the derived class")
+
     def similar_items(self, query_item: Any, top_k: int = 10) -> List[Any]:
         """
         Find similar items for a list of query items.

rtrec/models/slim.py

@@ -26,7 +26,6 @@ def fit(self, user_interactions: Iterable[Tuple[Any, Any, float, float]], update
         interaction_matrix = self.interactions.to_csc(item_ids)
         self.model.partial_fit_items(interaction_matrix, item_ids)
 
-    @override
     def _bulk_fit(self, interaction_matrix: csc_matrix) -> None:
         """
         Fit the recommender model on the given interaction matrix.
@@ -46,6 +45,9 @@ def _recommend(self, user_id: int, candidate_item_ids: List[int], top_k: int = 1
         interaction_matrix = self.interactions.to_csr(select_users=[user_id])
         return self.model.recommend(user_id, interaction_matrix, candidate_item_ids, top_k=top_k, filter_interacted=filter_interacted)
 
+    def _recommend_batch(self, user_id, interaction_matrix, candidate_item_ids = None, top_k = 10, filter_interacted = True):
+        return self.model.recommend(user_id, interaction_matrix, candidate_item_ids, top_k=top_k, filter_interacted=filter_interacted)
+
     def _similar_items(self, query_item_id: int, top_k: int = 10) -> List[int]:
         """
         Find similar items for a list of query items.

rtrec/recommender.py

@@ -92,7 +92,7 @@ def recommend_batch(self, users: List[Any], top_k: int = 10, filter_interacted:
         :param filter_interacted: Whether to filter out items the user has already interacted with
         :return: List of top-K item indices recommended for each user
         """
-        return [self.recommend(user, top_k, filter_interacted) for user in users]
+        return self.model.recommend_batch(users, top_k, filter_interacted)
 
     def similar_items(self, query_items: List[Any], top_k: int = 10) -> List[List[Any]]:
         """