PAC-Bayesian Bound for the Conditional Value at Risk

The goal in statistical learning is to learn hypotheses that generalize well, which is typically formalized by seeking to minimize the expected risk associated with a given loss function.