Repository navigation
Broadcasting error in avg_score #193
Description
I get the following error with avg_score multilabel strategy -
File "C:\Users\localuserSKSG\anaconda3\envs\alenv\lib\site-packages\modAL\models\base.py", line 175, in query
query_result, query_metrics = self.query_strategy( File "C:\Users\localuserSKSG\anaconda3\envs\alenv\lib\site-packages\modAL\multilabel.py",
line 258, in avg_score classwise_scores = classwise_confidence*(classwise_predictions-1/2)
ValueError: operands could not be broadcast together with shapes (26872,2,3) (26872,3)
During handling of the above exception, another exception occurred:
Traceback (most recent call last): File "avg-score.py", line 264, in
query_index, query_instance = learner.query(X_pool_np, n_instances=n_instances)
File "C:\Users\localuserSKSG\anaconda3\envs\alenv\lib\site-packages\modAL\models\base.py", line 180, in query
query_result = self.query_strategy( File "C:\Users\localuserSKSG\anaconda3\envs\alenv\lib\site-packages\modAL\multilabel.py", line 258, in
avg_score classwise_scores = classwise_confidence*(classwise_predictions-1/2) ValueError: operands could not be broadcast together with shapes (26872,2,3) (26872,3)
We do not see this error with avg_confidence. Clearly this is because of these lines of code in avg_score which are not present in avg_confidence-
classwise_predictions = classifier.predict(X_pool)
classwise_scores = classwise_confidence*(classwise_predictions-1/2)
Does anyone has a solution/workaround for this?
Activity
naveedyassin commentedon Mar 22, 2025
The error message indicates that the classwise_confidence and classwise_predictions - 1/2 arrays have incompatible shapes and cannot be broadcast together.
Understanding the Issue:
• classwise_confidence has shape (26872, 2, 3).
• classwise_predictions - 1/2 has shape (26872, 3).
• Since they do not match, NumPy cannot perform element-wise multiplication.
Possible Fix:
1. Check classifier.predict(X_pool) Output:
• The shape mismatch suggests that classifier.predict(X_pool) returns (26872, 3), while classwise_confidence has an extra dimension (26872, 2, 3).
• Ensure that classifier.predict(X_pool) returns the expected number of class probabilities. If it is returning a (26872, 3), confirm why classwise_confidence has (26872, 2, 3).