Bounds on the Jensen Gap, and Implications for Mean-Concentrated Distributions

Abstract

This paper gives upper and lower bounds on the gap in Jensen's inequality, i.e., the difference between the expected value of a function of a random variable and the value of the function at the expected value of the random variable. The bounds depend only on growth properties of the function and specific moments of the random variable. The bounds are particularly useful for distributions that are concentrated around the mean, a commonly occurring scenario such as the average of i.i.d. samples and in statistical mechanics.

Cited by in corpus (6)

• Disentangled Contrastive Collaborative Filtering
• Towards Error Measures which Influence a Learners Inductive Bias to the Ground Truth
• Loss function based second-order Jensen inequality and its application to particle variational inference
• Sum-Rate Maximization of Multicell MISO Networks with Limited Information Exchange
• A convex analysis approach to tight expectation inequalities
• Contrastive Representation Learning with Trainable Augmentation Channel