This letter provides a simple extension of boosting methods for binary data where the probability of mislabeling depends on the label of an example. Loss functions are derived from the statistical perspective, which is based on likelihood analysis. Our proposed methods can be interpreted as a...