Classifier with Noisy Annotators

You are given a dataset where several individuals (annotators) have each provided labels for a set of samples. The labels are stored in a 2‑D integer array annotations: annotations[i][j] is the class label given by annotator j for sample i. All labels are integers from 0 to num_classes - 1. Some annotators may be unreliable. To detect them, we first compute a “gold” label for each sample using a simple majority vote among all annotators. If multiple classes tie for the…