Consistency versus Realizable H-Consistency for Multiclass Classification
By Philip M. Long, Rocco A. Servedio. 2013
58 citations recorded by the source corpus.
Subfields: Evaluation & Benchmarks
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Evaluation & Benchmarks
- Anton Osokin
- Simon Lacoste-Julien
- Corinna Cortes
- Giulia DeSalvo
- Mehryar Mohri
- Kirill Struminsky
- Yarin Gal
- Nicolas Papernot
Citations in the corpus
- Quantifying Learning Guarantees for Convex but Inconsistent Surrogates (2018, background): “There have indeed been several works to analyze inconsistent surrogates [12, 18, 5, 14], but they usually end the story with proving that some surrogate (or a family or surrogates) is not consistent.”
- Cross-Entropy Loss Functions: Theoretical Analysis and Applications (2023, background): “Another line of research focuses on realizable H-consistency guarantees (Long & Servedio, 2013; Zhang & Agarwal, 2020; Kuznetsov et al., 2014), which provides hypothesis set-specific consistency guarantees under the assumption that the underlying distribution is H-realizable.”
- Learning with Rejection (2016, background): “Long and Servedio [23] proposed instead a notion of H-consistency precisely meant to take the hypothesis set used into consideration.”