Weighted majority vote

Trust some voters more

In a plain majority vote every classifier counts equally. A weighted vote lets stronger classifiers count more, for example by giving each one a weight based on its validation accuracy.

Implement weighted_majority_vote(predictions, weights).

• predictions has shape (m, n): m classifiers voting on n samples, with integer labels 0, 1, 2, ….
• weights has length m and must be nonnegative.
• For each sample, add each classifier's weight to the total for the label it predicts. The winning label is the one with the largest total.
• Ties go to the smaller label.

Return an integer array of length n.

Constraints

• Negative weights raise ValueError.
• If the weights sum to zero, raise ValueError.
• len(weights) must equal the number of rows in predictions, otherwise raise ValueError.
• Do not modify the inputs.

Hints