Fast Rate Generalization Error Bounds: Variations on a Theme

Wu, Xuetong; Manton, Jonathan H.; Aickelin, Uwe; Zhu, Jingge

doi:10.1109/itw54588.2022.9965761

Cited by 6 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To address this limitation, several works, inspired by some PAC-Bayesian literature, have proposed fast-rate information-theoretic bounds [24,25,64], demonstrating a good characterization in some instances of non-convex settings such as deep learning. More recently, [65,66,71] present IOMI bounds for the Gaussian mean estimation problem, that achieve optimal convergence rates, contrasting previous bounds [9,70].…”

Section: Introductionmentioning

confidence: 94%

“…Recently, [65,66] use the unexpected excess risk as the DV auxiliary function and invoke the (η, c)central condition to establish some optimal-rate bounds for specific learning problems, e.g., Gaussian mean estimation. This again highlights the significance of selecting appropriate DV auxiliary functions and corresponding assumptions tailored to different learning problems.…”

Section: Extensionsmentioning

confidence: 99%

See 1 more Smart Citation

Perspectives from an Endocrinologist at the Ottawa Hospital - An Interview with Dr. Christopher Tran

Wang

Phuong²

2020

UOJM

View full text Add to dashboard Cite

Dr. Christopher Tran is an endocrinologist at The Ottawa Hospital known for his work on medical education and eConsults. He is also a friendly, familiar face among second-year uOttawa medical students during their Endocrinology Block. We sat down with Dr. Tran to talk about his current work as well as his perspectives on research and social media use among academics. Dr. Tran also provides advice to medical students with regards to deciding on a medical specialty to pursue and how to approach the CaRMS application process.

show abstract

Section: Introductionmentioning

confidence: 94%

Section: Extensionsmentioning

confidence: 99%

Perspectives from an Endocrinologist at the Ottawa Hospital - An Interview with Dr. Christopher Tran

Wang

Phuong²

2020

UOJM

View full text Add to dashboard Cite

show abstract

“…Subsequent to our work, Wu et al [2022] prove fast-rate conditional generalization bounds based on the (𝑣, 𝜂)-central assumption, which is weaker than the Bernstein condition and captures unbounded and subgaussian losses.…”

Section: Compression Scheme Priorsmentioning

confidence: 94%

Algorithms and frameworks for preventing privacy leakage and overfitting in machine learning

Zakynthinou

View full text Add to dashboard Cite

Machine learning algorithms aim to learn useful models for the population, only having access to a training dataset as its proxy. However, they may overly tailor their output to the training dataset, which leads to the following two types of unwanted behavior: 1. they may leak too much information about specific datapoints included in the training dataset, violating the privacy of the individuals who participate in it, or 2. they may "overfit" their training dataset, that is, their empirical performance on the training dataset does not generalize well to their true performance on the population. This thesis contributes tools for preventing or reasoning about these two unwanted behaviors.

show abstract

Exactly Tight Information-Theoretic Generalization Error Bound for the Quadratic Gaussian Problem

Zhou,

Tian,

Liu

2023

2023 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

Fast Rate Generalization Error Bounds: Variations on a Theme

Cited by 6 publications

References 10 publications

Perspectives from an Endocrinologist at the Ottawa Hospital - An Interview with Dr. Christopher Tran

Perspectives from an Endocrinologist at the Ottawa Hospital - An Interview with Dr. Christopher Tran

Algorithms and frameworks for preventing privacy leakage and overfitting in machine learning

Exactly Tight Information-Theoretic Generalization Error Bound for the Quadratic Gaussian Problem

Contact Info

Product

Resources

About