Reset random number generator for reproducibility

cornac/data/dataset.py

@@ -426,7 +426,13 @@ def from_uirt(cls, data, seed=None):
         """
         return cls.build(data, fmt="UIRT", seed=seed)
 
+    def reset(self):
+        """Reset the random number generator for reproducibility"""
+        self.rng = get_rng(self.seed)
+        return self
+
     def num_batches(self, batch_size):
+        """Estimate number of batches per epoch"""
         return estimate_batches(len(self.uir_tuple[0]), batch_size)
 
     def idx_iter(self, idx_range, batch_size=1, shuffle=False):

cornac/eval_methods/base_method.py

@@ -28,6 +28,7 @@
 from ..metrics import RatingMetric
 from ..metrics import RankingMetric
 from ..experiment.result import Result
+from ..utils import get_rng
 
 
 def rating_eval(model, metrics, test_set, user_based=False):
@@ -239,6 +240,7 @@ def __init__(
         self.exclude_unknowns = exclude_unknowns
         self.verbose = verbose
         self.seed = seed
+        self.rng = get_rng(seed)
         self.global_uid_map = OrderedDict()
         self.global_iid_map = OrderedDict()
 
@@ -362,6 +364,11 @@ def sentiment(self, input_modality):
             )
         self.__sentiment = input_modality
 
+    def _reset(self):
+        """Reset the random number generator for reproducibility"""
+        self.rng = get_rng(self.seed)
+        self.test_set = self.test_set.reset()
+
     def _organize_metrics(self, metrics):
         """Organize metrics according to their types (rating or raking)
 
@@ -559,12 +566,13 @@ def evaluate(self, model, metrics, user_based, show_validation=True):
         res: :obj:`cornac.experiment.Result`
 
         """
-        self._organize_metrics(metrics)
-
         if self.train_set is None:
             raise ValueError("train_set is required but None!")
         if self.test_set is None:
             raise ValueError("test_set is required but None!")
+
+        self._reset()
+        self._organize_metrics(metrics)
 
         ###########
         # FITTING #

cornac/eval_methods/cross_validation.py

@@ -81,17 +81,14 @@ def __init__(
 
     def _partition_data(self):
         """Partition ratings into n_folds"""
-
-        rng = get_rng(self.seed)
-
         fold_size = int(self.n_ratings / self.n_folds)
         remain_size = self.n_ratings - fold_size * self.n_folds
 
         partition = np.repeat(np.arange(self.n_folds), fold_size)
-        rng.shuffle(partition)
+        self.rng.shuffle(partition)
 
         if remain_size > 0:
-            remain_partition = rng.choice(
+            remain_partition = self.rng.choice(
                 self.n_folds, size=remain_size, replace=True, p=None
             )
             partition = np.concatenate((partition, remain_partition))

cornac/eval_methods/ratio_split.py

@@ -93,7 +93,7 @@ def validate_size(val_size, test_size, num_ratings):
         return int(train_size), int(val_size), int(test_size)
 
     def _split(self):
-        data_idx = get_rng(self.seed).permutation(len(self._data))
+        data_idx = self.rng.permutation(len(self._data))
         train_idx = data_idx[:self.train_size]
         test_idx = data_idx[-self.test_size:]
         val_idx = data_idx[self.train_size:-self.test_size]

cornac/models/recommender.py

@@ -96,8 +96,8 @@ def fit(self, train_set, val_set=None):
         self : object
         """
         self.reset_info()
-        self.train_set = train_set
-        self.val_set = val_set
+        self.train_set = train_set.reset()
+        self.val_set = None if val_set is None else val_set.reset()
         return self
 
     def score(self, user_idx, item_idx=None):