Generalization Bounds for Meta-Learning via PAC-Bayes and Uniform Stability

Neural Information Processing Systems 

We employ bounds for uniformly stable algorithms at the base level and bounds from the P AC-Bayes framework at the meta level.