Relative deviation learning bounds and generalization with unbounded loss functions

By Corinna Cortes, Spencer Greenberg, Mehryar Mohri. 2013

40 citations recorded by the source corpus.

We present an extensive analysis of relative deviation bounds, including detailed proofs of two-sided inequalities and their implications. We also give detailed proofs of two-sided generalization bounds that hold in the general case of unbounded loss functions, under the assumption that a moment of the loss is bounded. We then illustrate how to apply these results in a sample application: the analysis of importance weighting.

Subfields: Artificial Intelligence, Computational Mechanics, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Mehryar Mohri
2. Ananda Theertha Suresh
3. Corinna Cortes
4. Ningshan Zhang
5. Andrés Muñoz Medina
6. Wouter M. Kouw
7. Virginia Smith
8. David A. Sontag

In Computational Mechanics

1. Mehryar Mohri
2. Corinna Cortes
3. Ningshan Zhang
4. Ananda Theertha Suresh
5. Wouter M. Kouw
6. Andrés Muñoz Medina
7. Vitaly Kuznetsov
8. Afshin Rostamizadeh

In Statistics and Probability

1. Corinna Cortes
2. Mehryar Mohri
3. Ananda Theertha Suresh
4. Ningshan Zhang

Citations in the corpus

• Multiple-Source Adaptation with Domain Classifiers (2020, background): “…δ > 0 , with probability at least 1 δ , for any α ≥ 1 , each of the following two inequalities holds, Theorem 9. where To prove Theorem 9, we need the following relative deviation bound (Corollary 7) from Cortes et al. [2019a], which is adapted to the learning scenario of kernel density estimation.”