Merge pull request #22 from RUCAIBox/master

Update

evaluator/evaluator.py

@@ -11,8 +11,8 @@
 metric_name = {metric.lower() : metric for metric in ['Hit', 'Recall', 'MRR', 'AUC', 'Precision', 'NDCG']}
 
 # These metrics are typical in topk recommendations
-topk_metric = {'hit', 'recall', 'precision', 'ndcg'}
-other_metric = {'auc', 'mrr'}
+topk_metric = {'hit', 'recall', 'precision', 'ndcg', 'mrr'}
+other_metric = {'auc'}
 
 class AbstractEvaluator(metaclass=abc.ABCMeta):
     """The abstract class of the evaluation module, its subclasses must implement their functions

utils/metrics.py

@@ -12,16 +12,17 @@ def hit(rank, label, k):
     """
     return int(any(rank[label] <= k))
 
-def mrr(rank, label, k=None):
+def mrr(rank, label, k):
     """The MRR (also known as mean reciprocal rank) is a statistic measure for evaluating any process that produces a list
     of possible responses to a sample of queries, ordered by probability of correctness. 
 
     url:https://en.wikipedia.org/wiki/Mean_reciprocal_rank
 
     """
     ground_truth_ranks = rank[label]
-    if ground_truth_ranks.all():
-        return (1 / ground_truth_ranks.min())
+    ground_truth_at_k = ground_truth_ranks[ground_truth_ranks <= k]
+    if ground_truth_at_k.shape[0] > 0:
+        return (1 / ground_truth_at_k.min())
     return 0
 
 def recall(rank, label, k):